Transportation Research: Proceedings of TPMDC 2022 (Lecture Notes in Civil Engineering, 434) 9819960894, 9789819960897

Citation preview

Lecture Notes in Civil Engineering

Dharamveer Singh · Avijit Maji · Omkar Karmarkar · Monik Gupta · Nagendra Rao Velaga · Solomon Debbarma   Editors

Transportation Research Proceedings of TPMDC 2022

Lecture Notes in Civil Engineering Volume 434

Series Editors Marco di Prisco, Politecnico di Milano, Milano, Italy Sheng-Hong Chen, School of Water Resources and Hydropower Engineering, Wuhan University, Wuhan, China Ioannis Vayas, Institute of Steel Structures, National Technical University of Athens, Athens, Greece Sanjay Kumar Shukla, School of Engineering, Edith Cowan University, Joondalup, WA, Australia Anuj Sharma, Iowa State University, Ames, IA, USA Nagesh Kumar, Department of Civil Engineering, Indian Institute of Science Bangalore, Bengaluru, Karnataka, India Chien Ming Wang, School of Civil Engineering, The University of Queensland, Brisbane, QLD, Australia Zhen-Dong Cui, China University of Mining and Technology, Xuzhou, China

Lecture Notes in Civil Engineering (LNCE) publishes the latest developments in Civil Engineering—quickly, informally and in top quality. Though original research reported in proceedings and post-proceedings represents the core of LNCE, edited volumes of exceptionally high quality and interest may also be considered for publication. Volumes published in LNCE embrace all aspects and subfields of, as well as new challenges in, Civil Engineering. Topics in the series include: • • • • • • • • • • • • • • •

Construction and Structural Mechanics Building Materials Concrete, Steel and Timber Structures Geotechnical Engineering Earthquake Engineering Coastal Engineering Ocean and Offshore Engineering; Ships and Floating Structures Hydraulics, Hydrology and Water Resources Engineering Environmental Engineering and Sustainability Structural Health and Monitoring Surveying and Geographical Information Systems Indoor Environments Transportation and Traffic Risk Analysis Safety and Security

To submit a proposal or request further information, please contact the appropriate Springer Editor: – Pierpaolo Riva at [email protected] (Europe and Americas); – Swati Meherishi at [email protected] (Asia—except China, Australia, and New Zealand); – Wayne Hu at [email protected] (China). All books in the series now indexed by Scopus and EI Compendex database!

Dharamveer Singh · Avijit Maji · Omkar Karmarkar · Monik Gupta · Nagendra Rao Velaga · Solomon Debbarma Editors

Transportation Research Proceedings of TPMDC 2022

Editors Dharamveer Singh Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

Avijit Maji Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

Omkar Karmarkar Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

Monik Gupta Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

Nagendra Rao Velaga Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

Solomon Debbarma Department of Civil Engineering Indian Institute of Technology Bombay Powai, Maharashtra, India

ISSN 2366-2557 ISSN 2366-2565 (electronic) Lecture Notes in Civil Engineering ISBN 978-981-99-6089-7 ISBN 978-981-99-6090-3 (eBook) https://doi.org/10.1007/978-981-99-6090-3 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore Paper in this product is recyclable.

Introduction

The practical problems associated with transportation systems engineering in many developing countries including India are different and complex. Therefore, researchers and agencies have been working to analyse the challenges and to identify implementable solutions for various transportation engineering related problems as per prevailing conditions. The earliest roots of the Conference on Transportation Planning and Implementation Methodologies for Developing Countries (TPMDC) can be traced back to the International Workshop/Conference series started by The Transportation Systems Engineering (TSE) group of IIT Bombay about 26 years ago by making the first announcement in the WCTRS (World Conference on Transport Research Society) Newsletter. The first Workshop of the series was organized in December 1994 on Impact Evaluation and Analysis of Transportation Projects in Developing Countries, IEATP-94. Riding on the success of the workshop, the second of the series, Transportation Planning, and Implementation Methodologies for Developing Countries (TPMDC-96) was conducted in December 1996. Starting from year 1996, international conference on TPMDC are conducted biennially. TPMDC 2022 brings an ideal platform for researchers, practitioners, and agencies to share and exchange the experience among the transportation professionals of the developing and developed nations.

TPMDC 2022 The 14th edition of TPMDC was held from 19th to 21st December 2022, at IIT Bombay, Mumbai. The conference covers the national and local level transport planning, traffic operation and management, pavement design, materials characterization, highway safety, geometric design, intelligent transportation systems and freight transport activities in India. The conference themes were designed to facilitate the contribution of research articles along the six following tracks.

v

vi

Introduction

Track I: Transport Modes: General Highways, Railways, Airways and Waterways This theme includes general aspects of all transport modes including air transport, rail-based transport, inland and international waterways, and non-motorized transport modes. The theme covers research on various topics such as air traffic control and management; port, harbor, and fleeting services; railway system planning, design, operation, and management; freight transportation and operations using rail, water, and air modes. Research on infrastructure planning and geometric design aspects of transportation projects are also included in this theme.

Track II: Pavement Systems Engineering This theme will focus on all aspects of material characterization, analysis, design, construction, evaluation, and maintenance of pavement structures. These aspects include broad areas of characterization of conventional and innovative pavement materials; pavement material modelling; pavement recycling; stabilization of pavement layers; analysis and design of bituminous, concrete, composite, and other kinds of pavements; life cycle cost analysis; pavement maintenance and management systems; pavement drainage; airport pavements and railway track.

Track III: Transportation Planning, Policy and Economics This theme includes research on theoretical and empirical aspects of travel demand and behavioral modelling and forecasting, network design for passenger and freight transport, planning of urban transport systems, and logistics planning. The theme covers several aspects of transportation economics such as public transport pricing, user impact of transport projects, cost benefit analysis and project evaluation, and transport as a means for economic development. The theme also includes research on transport policy analysis, and social equity in transport.

Track IV: Traffic Management, Operations, and Safety This theme deals with traffic aspects of highways and urban roads such as traffic flow theory and modelling, traffic control and management, transport network analysis, operations and management of passenger and freight traffic, and ICT for traffic systems. This theme also includes the various aspects of transport safety such as driver and infrastructural factors, externalities and policies for transport safety aspects,

Introduction

vii

vulnerable road user safety, safety in public transit, and other modes of transport. The theme also covers the topic on operations and management of public transport systems.

Track V: Emerging Transportation Technologies This theme focuses on all aspects of technological advances in transportation systems. The topics broadly include battery and electric operated vehicles, connected and autonomous vehicles; internet of things, V2X communication, and intelligent transportation systems; technology enabled models for mobility services of passenger and freight transport; technology enabled multi-modal integration; robotics, artificial intelligence, human-machine interfaces, computer vision, image processing, and augmented reality in transportation; big data analytics for transportation.

Track VI: Sustainable Mobility in Transportation This theme includes research related to environmental impact assessment of transport projects and mitigation strategies; various interactions between transport, health and associated policies; emissions from vehicles and other transport infrastructure; policies and planning for sustainability in transport systems; renewable energy applications in transport sector; pollution and environmental issues of all transport modes; strategies, technological interventions, policies and management of sustainable and socially inclusive transport systems.

Acknowledgements The financial aspect of the conference was primarily supported by our sponsors: Navi Mumbai Municipal Corporation (NMMC), Blacktop Infrastructure Private Limited, Thane Municipal Corporation, Maharashtra State Road Development Corporation (MSRDC), Bitpath Private Limited, Savi Infrastructures and Properties, Dineshchandra R. Agrawal Infracon, City and Industrial Development Corporation (CIDCO), Viatop Premium, AIC Infrastructures, Markolines, Zydex Industries, Samarth InfraENGG Technocrats, KABA Infratech, Yash Innovative Solutions, Bitumen Corporation, Mumbai Port Authority and Balajee Infratech & Constructions. We would like to acknowledge all the authors for submitting and presenting their recent quality work at TPMDC 2022. We are especially grateful to our delegates including key note speakers and invited speakers for sharing their valuable knowledge with the participants. We extend our appreciation to the renowned researchers who

viii

Introduction

supported us in the critical review process of the received papers. We thank the members of the conference organizing committee Dr. K. V. Krishna Rao, Dr. Tom V. Mathew, Dr. Gopal R. Patil, and Dr. P. Vedagiri, students of the transportation systems engineering and IIT Bombay administration for their valuable assistance for organizing the conference. Dr. Dharamveer Singh Dr. Avijit Maji Dr. Nagendra Rao Velaga Dr. Solomon Debbarma Mr. Omkar Karmarkar Mr. Monik Gupta

Contents

Pavement Systems Engineering Development of Maintenance Priority Index for Urban Road Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Saurabh Singh Yadav, Aakash Gupta, Sachin Gowda, and Yogesh Aggarwal Modelling of Deflection Basin Parameters of Asphalt Pavements Using Artificial Neural Networks and Adaptive Neuro-Fuzzy Inference Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sachin Gowda, K. Vaishakh, Aakash Gupta, R. Prakash, and G. Kavitha

3

19

Comparative Studies on Gel-Incorporated Flexible Pavement . . . . . . . . . . Delvin J. Joseph, Padmakumar Radhakrishnan, and Vignesh Dhurai

31

Performance Assessment of Premix Carpet for Low-Volume Roads . . . . Nishant Bhargava, Anjan Kumar Siddagangaiah, and Teiborlang Lyngdoh Ryntathiang

41

Development of Resilient Modulus Model for the Bituminous Course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Paras Markana, Bharath Gottumukkala, Akshay Gundla, Ambika Behl, and Tejaskumar Thaker Forensic Investigations for Failure of Flexible Pavements: A Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Abhishek Mittal and Amit Kumar Effect of Subgrade Stabilization on Pavement Design: Material Optimization and Economic Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sudeshna Purkayastha, Ritu Raj Patel, Veena Venudharan, and Ajitkumar Vadakkoot

53

65

75

ix

x

Contents

An Effective Bitumen-Friendly Polymer for Superior Roadway Performance and Durability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Krishna Srinivasan, Sachin Raje, and Deepak Madan

87

Rheological Investigation of Soft Grade Asphalt Binder Modified With Crumb Rubber-Nanosilica Composite . . . . . . . . . . . . . . . . . . . . . . . . . . Tabish Mehraj, Mohammad Shafi Mir, and Bijayananda Mohanty

99

Utilization of Waste Polyethylene in Open Graded Friction Course . . . . . 121 Debashish Kar, Mahabir Panda, and Subhashree Jena Rheological Characteristics of Waste Engine Oil-Modified Bituminous Binder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 T. Srikanth, R. Amal, and M. Sivakumar Rheological Investigation of Nano Silica-SBS Modified Soft Grade Bitumen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Kulsuma Salam, Mohammad Shafi Mir, and Bijayananda Mohanty Performance Evaluation of Conventional and Field-Produced High Modulus Asphalt Binders and Mixtures . . . . . . . . . . . . . . . . . . . . . . . . 167 S. Chinmaya, G. Bharath, Sarfaraz Ahmed, Ambika Behl, and Muttana S. Balreddy Influence of Feeding Sequence on the Fluidity and Fluidity Time of Cement Asphalt Mortar for High-Speed Rail Slab Track . . . . . . . . . . . . 181 Rahul Reddy Banapuram, Kranthi K. Kuna, and M. Amarnatha Reddy Stiffness and Cracking Resistance Evaluation of Cold Bitumen Emulsion Mixtures Incorporated with Waste Glass Aggregates . . . . . . . . 193 Mohammad Iqbal Malik, Mohammad Shafi Mir, Bijayananda Mohanty, and Mehnaza Akhter Effect of Warm Mix Additive, RAP, and Waste Oil on Rheological Properties of Binder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 Mayank Mehrotra, Abhishek Mittal, and Dipak Rathva Usage of Exclusive Plastic-Based Pre-fabricated Panels in Road Construction: A Feasibility Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221 Chandu Yadav Ullam, Gagandeep Singh, and S Shankar Evaluation of Effective Thickness of Bituminous Block Pavement Using Finite Element Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 Vignesh Dhurai, Padmakumar Radhakrishnan, and Delvin J. Joseph Effect of Cement and RAP Content on Full Depth Reclamation (FDR) of Low-Volume Roads: A Response Surfaced-Based Study . . . . . . 245 Supratim Kaushik, Anugu Sairaj, Vishal Kirti, and S. Anjan Kumar

Contents

xi

High- and Intermediate-Temperature Performance of Crumb Rubber-Nano-Alumina-Modified Asphalt Binder . . . . . . . . . . . . . . . . . . . . . 257 Zahir Iqbal, Mohammad Shafi Mir, and Bijayananda Mohanty Development of Pavement Maintenance Management System (PMMS) for an Urban Road in New Delhi, India . . . . . . . . . . . . . . . . . . . . . 277 K. Surya Kiran, Vidhi Vyas, G. Kavitha, Pradeep Kumar, and Ashok Kumar Sagar A Study on the Design of Sustainable Bituminous Concrete Pavement with Subgrade Soil Stabilization . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 Uppuluri Siva Rama Krishna, T. Vijaya Gowri, and E. Sree Keerthana Investigation of Effect of Aging on Reflection Cracking in Overlay . . . . . 301 Sarfaraz Ahmed, Abhishek Tiwari, and Vijay Kakade Ultraviolet Radiation Ageing of Asphalt: A Critical Review . . . . . . . . . . . . 313 Suhas Pandhwale, Adyasha Mohanty, Anush K. Chandrappa, and Vijayakrishna Kari Strength and Durability Characteristics of Stabilised Clayey Soil for Low Volume Roads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323 Samir Saurav and Sanjeev Sinha Transportation Planning, Policy and Economics Sustainability Integration Index of Metro and Buses for Evaluation of Transport Policies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341 Rohit Raghuwanshi, Madhu Errampalli, Minal Chandra, and Shubha Khatri A Critical Review of Strategic Decisions for Travel Time Performance Improvement of Public Transport System . . . . . . . . . . . . . . . 351 Narendra Dudhe, Pradeep Kumar Agarwal, and Amit Vishwakarma Trip Generation Based on Land Use Characteristics: A Review of the Techniques Used in Recent Years . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367 Saumya Anand, Pritikana Das, and G. R. Bivina Mode Choice Behaviour of Students Using Structural Equation Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379 Farahna Amin, Devika Babu, and M. V. L. R. Anjaneyulu Pune Metro: Preference Survey and Analysis for the Modal Split of Work Trips to Hinjewadi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391 Priya Hirave and Vidula Sohoni Passengers Perception and Satisfaction Level Towards Public Transport: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403 Arjun Singh Lodhi and Anuj Jaiswal

xii

Contents

Effect of Land Use Pattern on Bus Blockage Duration at Curb Side Bus Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411 Remya K. Padinjarapat, Darshana Othayoth, K. V. Krishna Rao, and Tom V. Mathew Spatio-Temporal Factors Affecting Short-Term Public Transit Passenger Demand Prediction: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421 K. Shanthappa Nithin and Raviraj H. Mulangi Assessment of Satisfaction Level for Bus Transit Systems in Bhopal . . . . 431 Anuj Jaiswal, Siddharth Rokade, and Neelima C. Vijay Strategies for Improving Travel Time Performance of Multimodal Transport System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449 P. K. Agarwal, R. Tanwar, and A. Jain Traffic Management, Operations, and Safety A Review on Pedestrian Level of Service for Sidewalks . . . . . . . . . . . . . . . . 463 N. C. Vijay, S. Rokade, and G. R. Bivina Assessment of Pedestrian Safety at Urban Uncontrolled Intersections Using Surrogate Safety Measures: A Case Study of Bhopal City . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475 Dungar Singh, Pritikana Das, and Vasu Verma A Review of Safety and Operational Impacts of Various Speed Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487 Abhinav Mishra, A. Mohan Rao, and Darshana Othayoth Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499 P. Vijayalakshmi, Nitin Kumar, and Vivek R. Das Capacity Analysis and Safety Assessment of Unsignalized Intersection Using Conflict Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509 P. H. Souparnika, Sheela Alex, and Padmakumar Radhakrishnan A Statistical Approach to Estimate Gap Acceptance Parameter at Three-Legged Uncontrolled Intersection . . . . . . . . . . . . . . . . . . . . . . . . . . 523 Khushbu Bhatt and Jiten Shah Identification of Infrastructural Causative Factors for Road Accidents on Urban Arterial Roads: A Case Study of Ahmedabad . . . . . 541 Poojan Pasawala and Bhavin Shah Modelling Longitudinal and Lateral Vehicle Movement Behavior Under Multiple Influencing Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555 Dhiraj Kinkar, Madhu Errampalli, Mukti Advani, and Saraswathi Sethi

Contents

xiii

Comprehensive Analysis of Road Accidents and Surrogate Measures to Enhance Road Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567 B. S. Jisha and M. Satyakumar Visualising Blackspot Improvement at Nagpur . . . . . . . . . . . . . . . . . . . . . . . 583 Raghav Chawla, S. Velmurugan, Mukti Advani, and K. Ravinder Proactive Safety Assessment at Unsignalized T-Intersection Using Surrogate Safety Measures: A Case Study of Bhopal City . . . . . . . . . . . . . 597 C. Noor Mohammed Parvez, Pritikana Das, and Dungar Singh Ranking-Based Methodology for Prioritization of Critical Pedestrian Infrastructure in and Around the Market Area: A Case Study of Aminabad Market, Lucknow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609 Haroon Rasheed Khan, Mokaddes Ali Ahmed, and Manish Dutta Machine Learning Categorization Algorithms for Traffic Conflict Ratings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623 Pushkin Kachroo, Anamika Yadav, Ankit Kathuria, Shaurya Agarwal, and Mahmudul Islam A Fuzzy Logic Approach on Pedestrian Crossing Behaviour at Unsignalized Intersection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631 M. Manoj, Vivek R. Das, and Nitin Kumar A Review on Surrogate Safety Measures Using Extreme Value Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641 Dungar Singh, Pritikana Das, and Indrajit Ghosh Examining the Evasive Behaviour of Pedestrians to Measure Their Degree of Vulnerabilities at Unsignalised Intersections . . . . . . . . . . . . . . . . 653 George Kennedy Lyngdoh, Aakash Bhardwaj, Manish Dutta, and Suprava Jena Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network Partitions Facilitating Feedback Linearization-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663 Saumya Gupta, Pushkin Kachroo, Shaurya Agarwal, and Kaan Ozbay Meta-Analysis of the Methodologies Used for Road Accident Costing and Conceptualizing Framework for Road Accident Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673 Adil Ata Azmi and Sewa Ram Investigating Pedestrian Crash Risk at Unsignalized Midblock Crosswalks on Arterial Road . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 695 Shubham Thapliyal, Heikham Pritam Singh, and RB. Sharmila Economic Benefit Assessment of Black Spot Improvements . . . . . . . . . . . . 705 Malolan Balaji, S. Velmurugan, and S. Padma

xiv

Contents

Emerging Technology, Logistics and Sustainability Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 725 V. A. Ajay Swaroog, Sheela Alex, and Padmakumar Radhakrishnan Quality Assessment of App-Based Bike Taxi Services by Benchmarking and Numerical Rating Approach: Guwahati . . . . . . . . 735 Lalit Swami, Mokaddes Ali Ahmed, and Suprava Jena Car Parking in Indian Cities: A Review of the Impediments to Sustainable Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 747 Paulose N. Kuriakose and Binayak Choudhury Limits to Commute: The Case of Indian Women . . . . . . . . . . . . . . . . . . . . . 763 Nachiket Gosavi and Naga Siva Gayatri Dittakavi Regularization of Micro-mobility Modes in an Emerging Economy: A Case of India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 777 Shalini Kumari Joint Mode and Shipment Size Choice for Interregional Transportation of Fruits and Vegetables in India . . . . . . . . . . . . . . . . . . . . . 789 K. A. Gayathri, V. Ansu, and M. V. L. R. Anjaneyulu Model for Estimating On-Street Night Parking Demand: A Case Study at Roorkee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 797 Ashwani Bokadia, Mokaddes Ali Ahmed, and Prasenjit Das A Framework on Understanding the Barriers of Smart Cities and Intelligent Transportation System in India . . . . . . . . . . . . . . . . . . . . . . . 809 M. B. Sushma, Y. Rashmitha, and Sandeepan Roy Stepping Towards Environment Friendly Roads: Government-Initiatives in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 821 Yogesh Manoj Biyani A Review of Berth Allocation Problem in Bulk Terminals . . . . . . . . . . . . . 831 Adnan Pasha and Rajat Rastogi MCDA Approach to Evaluate Planning and Consolidation Freight Strategies for Sustainable Goods Distribution: Case of Indian City Jaipur . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 843 Pankaj Kant, Sanjay Gupta, and Ish Kumar Evaluating the Parking Characteristics and Parking Demand of On-Street Parking in Silchar City . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 855 Prasenjit Das, Ashwani Bokadia, and Mokaddes Ali Ahmed

About the Editors

Prof. Dharamveer Singh is an associate professor at Department of Civil Engineering at Indian Institute of Technology (IIT) Bombay in India. Prof. Singh has been a faculty member at IIT Bombay since 2012. He received BE from MBM Engg. College Jodhpur, M.Tech. from IIT Kharagpur, and Ph.D. from University of Oklahoma, USA. Professor Singh’s field of specialization is in pavement engineering, focusing on recycling, stabilization, new and innovative technology for road construction, pavement design, and maintenance and preservation of pavements, forensic investigation. Professor Singh is closely associated with industry and highway fraternity on various fronts including pavement design, construction, training, and implementation of new technologies. Professor Singh is member of IRC committees on pavement design, low volume roads, asset management, and composite pavements. Also, Prof. Singh is associated with Bureau of Indian Standards (BIS), technical committee on Bitumen, Tar and Related Products, PCD 06. Professor Singh has been conferred with Pt. Jawaharlal Nehru Birth Centenary Award-2018 by Indian Roads Congress Nov. 2019. Professor Singh has published more than 100 papers in international journal and conferences. Prof. Singh is an Associate Editor for International Journal, Innovative Infrastructure solutions, Springer Publisher, and Editorial Board Member of International Journal of Pavement Research and Technology, and International Journal of Road Materials and Pavement Design. Dr. Avijit Maji is a professor in the Department of Civil Engineering, IIT Bombay, India. He received Doctor of Engineering from Morgan State University, USA in 2008; M.Tech. from IIT Kanpur in 2004; and B.E. from Bengal Engineering College (now known as Indian Institute of Engineering Science and Technology) in 2000. He has more than twenty years of academia and industry experience working for organizations in India and USA. At present, he is the Co-Chair of PerformanceBased Approaches and Applications Subcommittee of the Transportation Research Board (TRB), USA and Executive Secretary of the Transportation Research Group xv

xvi

About the Editors

of India. His area of expertise includes design, operations, and safety of transportation infrastructure. He has published more than eighty technical research papers in journals, edited books, and conferences. For his contribution, Maryland Society of Professional Engineers conferred him with the “Young Engineer of the Year Award” in 2011. He is a licensed Professional Engineer (PE) in Maryland, USA and Transportation Professional Certification Board’s certified Professional Traffic Operations Engineer (PTOE). Mr. Omkar Karmarkar is a senior Ph.D. fellow in Transportation Systems Engineering lab at IIT Bombay. He has completed his Bachelor of Civil Engineering from IIT Bombay. He is a recipient of prestigious Prime Ministers’ Research Fellowship given by the Ministry of Education. His area of research includes urban growth modelling, transportation planning, public transport scheduling, and transport policy. Mr. Monik Gupta is a PMRF research scholar in Transportation Systems Engineering at IIT Bombay. He has also served as construction manager in the TATA Steel for two years after completing his graduation from IIT Hyderabad in Civil Engineering. During his research at IIT Bombay, he is working on analyzing the impact of distracted driving under time pressure driving situation among motorized two-wheeler (MTW) riders. Prof. Nagendra Rao Velaga is working as a professor in the Transportation Systems Engineering Group of IIT Bombay. His major areas of interest include intelligent transportation systems (ITS), human factors in ITS and transport accessibility & mobility. He received his Ph.D. in Transport Studies from Loughborough University, UK. Before joining with IIT Bombay, he worked as a research fellow for 3 years at Research Councils UK (RCUK) digital economy research hub, University of Aberdeen, UK. Dr. velaga has about 15 years of experience in industry and academia in India and abroad. He has published more than eighty technical research papers in peer reviewed international journals. Dr. Solomon Debbarma is an assistant professor in Transportation Systems Engineering, Department of Civil Engineering at IIT Bombay. Before joining IIT Bombay, Debbarma was an adjunct lecturer and a postdoctoral research associate at the Ingram School of Engineering of Texas State University, San Marcos, USA. He received his Ph.D. and M.Tech. in Civil Engineering from IIT Roorkee. He graduated with a Civil Engineering degree from NIT Agartala. His primary research interests include rigid pavements, sustainable concrete materials (e.g., recycled aggregates, industrial wastes, agricultural wastes), and special concrete for pavement applications. He has more than 20 peer-reviewed publications and was the lead author of 13 of the papers published in peer-reviewed journals. He is Volunteer Member of the ACI Committee 555 on Recycled Concrete Aggregates and Young Member of

About the Editors

xvii

RILEM. Debbarma served as a reviewer for several international peer-reviewed journals. He is a Young Faculty Award recipient awarded by IIT Bombay. In addition, he was the Entrepreneur Lead for the NSF Innovation-Corps project: High-performance Cementitious Materials, funded by the National Science Foundation, USA in 2021.

Pavement Systems Engineering

Development of Maintenance Priority Index for Urban Road Network Saurabh Singh Yadav, Aakash Gupta, Sachin Gowda, and Yogesh Aggarwal

Abstract The estimated service life of a pavement is determined by design criteria such as geological considerations, water table movements, structural variations, and existing circumstances such as traffic intensity, drainage, and climate. The analysis of deformations and other variables, which influencing the pavement life, is a difficult task since the events that cause them are unpredictable and random in nature. It is unavoidable yet; these variables have an impact on the quality standards of the road network, resulting in decreased in their usable life. As a result, in order to remedy difficulties, it is important to assess or diagnose the current pavement conditions, both structurally and functionally. As a result, the issue of pavement evaluation, which deals with the mentioned element, is critical for pavement management. The functional testing on road surfaces, as assessed by its strength and durability during its service life, is dependent on several subjective measures of its stiffness and roughness. Structural Evaluation of Pavements is required to measure the structural strength of various layers of pavement. It also helps in evaluating a pavement’s remaining life and the thickness of overlay necessary. In the current study, a maintenance priority index has been developed using functional and structural parameters and also, it has been compared with the already available maintenance priority tools. Keywords Functional parameter · Structural parameter · Prioritization index

S. Singh Yadav · Y. Aggarwal National Institute of Technology, Kurukshetra, Haryana, India e-mail: [email protected] Y. Aggarwal e-mail: [email protected] A. Gupta (B) · S. Gowda CSIR-Central Road Research Institute, New Delhi, India e-mail: [email protected] S. Gowda e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_1

3

4

S. Singh Yadav et al.

1 Introduction Pavement distress and their continuous deterioration lead to poor riding comfort and safety concerns. Apart from pavement distresses, many other parameters affect the pavement quality. Since the events that lead to deformations and other factors that affect pavement life are unpredictable and random in nature, analysing them can be challenging. It is nonetheless inevitable that these factors will affect the road network’s quality standards and shorten its useful life. As a result, it is critical to evaluate or diagnose the current pavement conditions, both physically and functionally, in order to address issues. The topic of pavement evaluation, which addresses the aforementioned component, is therefore crucial for pavement management. The functional testing on road surfaces, as assessed by its strength and durability during its service life, is dependent on several subjective measures of its stiffness and roughness.

2 Literature Review This section provides a thorough study of the literature on functional pavement and structural pavement of pavement. Discusses the research done by several experts in fieldwork of pavement inspection related to the different factors that influence pavement roughness, models for predicting pavement efficiency. It also contains indices, researchers came up with to prioritize paving latest techniques used to model the various components of pavement, i.e., functional and structural evaluation, pavement performance prediction models, resource allocation optimization and maintenance prioritization procedures. Biasnchini et al. (2013) determine the relative relevance of various forms of distress in determining pavement quality and finding significant characteristics for establishing maintenance plans. The main goal of the study conducted by Boyapati and Kumar [1] is to calculate Pavement Condition Index (PCI) by collecting and analysing field data in order to priorities pavement maintenance [1]. The main aim of the study conducted by Al-Neami et al. is to evaluation of PCI using GIS data and Paver 5.2 software [2]. In Murshan and Vala [3], the primary goal of this research is to offer a maintenance priority index. In Mohammad et al. [4], the main motive of this research is to regulate a physical survey of 50 lane roads and to conduct a relative study on Pavement Condition Rating Methods (PCRM). In Janani et al. [5], the goal of this study is to provide a new approach for prioritizing pavement maintenance sections based only on the functional parameter of the pavement. The purpose of the study conducted by Afifyet et al. (2020) is to evolve an accurate IRI forecast model for flexible pavements using both (ANN) Artificial neural networks and Multiple Linear Regression. In Zhang et al. [6], this study’s overall goal was to evolve a structural indicator, using the Falling Weight Deflectometer data. The structural study of the pavement was completed using Benkelman Beam deflection (BBD) to determine the pavement’s strength. In Amin et al. (2018) [7], in this research, only non-destructive test technique to analyse pavements using

Development of Maintenance Priority Index for Urban Road Network

5

FWD data is collected. Thereafter to analyse the structural proportions of the current pavement. KGPBACK software was used for back-calculation. In U. Shaha et al. (2013), pavement condition evaluation, which covers distress, roughness, friction, and structural parameter, is an important part of pavement design, rehabilitation, and maintenance. The majority of the cost-effective maintenance and rehabilitation (M&R) methods are produced using the Pavement Management System (PMS) [8]. The weights are determined by employing the Analytical Hierarchy. In Gupta et al. [9], the aim of this study is that development of rural road maintenance priority index is based on functional and structural condition of rural road network [9].

3 Data Collection In this current study, 48 km road stretch in Delhi NCR, each with length of 1 km and average width of 3.5 m, was selected to collect the details of distress, roughness, and pavement deflection data. Data were collected on site to develop various models that would help in the maintenance of the urban road network. Field data were collected based on functional parameters predominantly prevalent in urban sections of the Delhi NCR Road network. Road roughness, pavement deterioration such as cracks, patching, ruts, potholes are among the functional parameters evaluated, ranging in severity from low to severe. Falling Weight Deflectometer is used to evaluate the structural behaviour of the pavement as shown in Fig. 1.

3.1 Functional Evaluation Data The functional evaluation of 48 km of the chosen road section comprised gathering functional metrics of road surface distresses (cracks, spalling, potholes, ruts), and road roughness (IRI), which are mostly found in the selected road section.

FIELD DATA COLLECTION

Pavement Functional Evaluation

Measurement of Distress

Fig. 1 Flowchart of data collection

Measurement of Roughness(IRI)

Pavement Structural Evaluation

Measurement of Deflection

6

S. Singh Yadav et al.

(a) Network Survey Vehicle (NSV)

(b) Hawkeye Tool

Fig. 2 Network survey vehicle and Hawkeye tool

3.2 Structural Evaluation Data A structural evaluation of selected road sections in Delhi-NCR was performed. The structural factor, i.e., Pavement Deflection, which is detected using FWD is mainly used for this study.

4 Equipment Used for Analysis of Pavement 4.1 Network Survey Vehicle (NSV) Network Survey Vehicle (NSV) uses the most up-to-date survey methodologies. Tools for processing video images, laser, GPS, etc. Survey-Vehicle is utilized for the automated collecting road asset management data, pavement maintenance management system data, and inventory data. Surface characteristics of pavement include: Cracking, Rutting, Raveling Potholes, Roughness and Studies on road safety audits. Data collected are analysed using Hawkeye processing tool as shown in Fig. 2.

4.2 Falling Weight Deflectometer (FWD) Falling Weight Deflectometer (FWD) is an impulse loading instrument that applies a weight to the pavement and measures the deflected shape of the surface. The falling mass is allowed to descend vertically on a set of springs positioned over a circular loading plate that applies the impulse force. Displacement sensors are positioned at varying radial distances from the centre of the load plate to measure the curved shape of the attached surface. The FWD variants are commercially available in both trailer and vehicle mount configurations as shown in Fig. 3. A predetermined amount of

Development of Maintenance Priority Index for Urban Road Network

Falling Weight Deflectometer (FWD)

7

FWD Software

Fig. 3 Falling weight deflectometer (FWD)

weight is dropped from a specified height onto a series of springs/buffers installed thereon. At various radial positions, the associated peak load and vertical tip surface deflections are measured and recorded.

5 Methodology The aim of this study is to produce an overall pavement condition index that can be used for urban road maintenance and makes good use of available resources. For this, firstly the functional data and structural data have been collected. Pavement distress and road roughness are considered as functional parameters while surface deflection is considered as structural parameter. After collecting all the data, Distress Index, Roughness Index, and Deflection Index have been developed individually. This is followed by the Overall Functional Condition Index and the Overall Structural Condition Index. Once, the Overall Functional Condition Index and the Overall Structural Condition Index are calculated, by giving a suitable weightage to them, the Overall Pavement Condition Index is obtained (AHP and Expert Choice software has been used to give the weightage). With the help of the Overall Pavement Condition Index, road maintenance and their ranking can be done. This technique can be used for various types of roads such as state highways, district roads, and rural roads as shown in Fig. 4.

5.1 Development of Functional Condition Indices To calculate the combined pavement condition index, the following equations given in Table 1 have been used to determine the pavement distress index (PCI Distress).

8

S. Singh Yadav et al.

Selection of Road Network

Data Collection

Functional Data

Distress & Roughness

Development of Functional Condition Index

Structural Data

Deflection(D0)

Development of Structural Condition Index

Determination of weightage using AHP Technique & Expert Choice Software

Development of Overall Functional Condition Index

Development of Overall Structural Condition Index

Development of Overall Pavement Condition Index (UR-OPCI)

Fig. 4 Overall methodology

The threshold value for all distress has been set to 60, indicating that the pavement needs repairs [9]. The percentages of L, M, and H in Table 1 above represent the proportion of the observed pavement with a given distress among the different severity levels. The denominator numbers are the maximum allowable extent (MAE) for each severity level. Table 1 Equations used to determine individual distress index Distress

Low severity index (LSI)

Medium severity index (MSI)

High severity index (HSI)

Longitudinal cracking

100−40* (%L/25)

100−40* (%M/20)

100−40* (%H/10)

Transverse cracking

100−40* (%L/25)

100−40* (%M/20)

100−40* (%H/10)

Alligator cracking

100−40* (%L/50)

100−40* (%M/25)

100−40* (%H/15)

Potholes

100−40* (%L/50)

100−40* (%M/30)

100−40* (%H/10)

Patching

100−40* (%L/50)

100−40* (%M/15)

100−40* (%H/10)

Raveling

100−40* (%L/80)

100−40* (%M/60)

100−40* (%H/30)

Development of Maintenance Priority Index for Urban Road Network

9

5.2 Analytical Hierarchy Process (AHP) AHP is a rating tool used to assign weights to a number of distress factors to determine their relative importance. The creativity, knowledge, and talent of each individual distress are determined by the AHP. It arithmetically synthesizes the multiple choices or perceptions, checks the coherence of judgements to evaluate each choice, and finally arrives at a result that represents the associated problem statement. The questionnaires are made available to road engineers, scientists, and students to support their individual perception and to support their expertise and knowledge [9].

5.3 AHP Weightage Determination AHP is a systematic approach that can be used to make complicated decisions. With this method, the problem is first broken down within a hierarchical structure, which is then broken down into several parts. This can assess any complicated decision challenge that involves expert views and perspectives. The AHP analyses any problem with the help of each person’s creativity, expertise, and experience. It mathematically combines the various judgments or perceptions, checks the consistency of the judgments to assess each judgement, and then arrives at an output to represent the appropriate issue statement [9]. The analytical hierarchy process approach is used in this study to determine the weighting of several distress criteria that have a significant impact on the International Roughness Index Distress and surface deflection. For this purpose, a questionnaire is shown in the Annexure, a total of 125 questionnaires is distributed, out of which 96 were answered and used to determine the proportional importance of different types of distress. Expert Choice 11 software was used to examine the consistency ratio (CR) of 125. If a response’s consistency ratio is greater than 0.1, it has been discarded. The AHP Expert Choice 11 software is used to calculate the Weightage factor of different distress, responses as shown in Table 2. Table 2 Weightage determined using AHP expert choice 11 software

Pavement distress categories

Weightage (Wi )

Longitudinal cracking (LC)

0.2

Transverse cracking (TC)

0.19

Alligator cracking (AC)

0.39

Potholes (PH)

0.14

Raveling (RA)

0.05

Patching (PA)

0.03

10

S. Singh Yadav et al.

5.4 Combined (PCI Distress) After assessing each stress index, the Combined PCI stress is calculated using the following Eq. 1. Because the impact of all stresses was not equal, stresses are weighted differently based on the expert choice 11 software to account for their individual impact, which is listed in Table 1. (PCI) Distress = 100 ∗ [1 − { 1 − LC/100} ∗ 0.20 ] ∗ [1 − { 1 − TC/100} ∗ 0.19 ] ∗ [1 − { 1 − AC/100} ∗ 0.39 ] ∗ [1 − { 1 − PA/100} ∗ 0.03 ] ∗ [1 − { 1 − RA/100} ∗ 0.05 ] ∗ [1 − { 1 − PH/100} ∗ 0.14 ]

(1)

5.5 Roughness Index (PCI Roughness) A correlation between the IRI (International Roughness Index, m/km) and the Ride Quality Rating (RQR) was discovered by means of regression analysis. The PCI (roughness) is calculated using the polynomials Eq. 2, which gives a perfect result [9]. (PCI) Roughness = 1.227 ∗ IRI2 − 17.73 ∗ IRI + 100

(2)

5.6 Development of Structural Condition Index (SCI) Structural Condition Index (SCI) is calculated based on its structure number, which is determined by the rebound deflection of the pavement surface, the layer coefficient and the thickness component of each pavement layer. (SCI) structural condition index is the ratio of Modified Structural Condition and Effective Structural Condition [9]. MSN = SN + 3.51 log10 (CBR) − 0.85(log10 CBR)2 − 1.43

(3)

Here, SN = 0.0393

n 

(an ∗ dn )

(4)

i=1

where “an ” represents the layer coefficients of n layers and “d n ” represents the thickness of n layers of pavement in mm. “an ” = 0.3 for Bituminous layer and 0.14 for GSB/subgrade layer. CBR—denotes the California Bearing Ratio of the pavement sub grade (%)

Development of Maintenance Priority Index for Urban Road Network

11

SCI = (MSN/SN effective) ∗ 100

(5)

Here, SN effective = 3.2 ∗ (Characteristic deflection in mm using FWD) − 0.63

(6)

5.7 Evaluation of Overall Functional Condition Index (OFCI) and Overall Structural Condition Index (OSCI) The Overall Functional Condition Index (OFCI) is influenced by the Functional Condition Index, which includes PCI (distress) and PCI (roughness), while the Overall Structural Condition Index (OSCI) is influenced by the SCI (deflection), respectively. The questionnaire survey with 125 answers, which corresponded to the questionnaire in the Annexure, will be used to determine the relative importance of OFCI and OSCI by using the Expert Choice 11 tool. The Weightage for the Pavement Distresses Index (PCI, Distresses) and the Roughness Index (PCI Roughness) are achieved with 65% and 35%, respectively. Whereas SCI (deflection) is achieved 100% Weightage. As a result, (OFCI) and (OSCI) were calculated by using Eqs. 7 and 8 OFCI = 0.65 ∗ PCI (distress) + 0.35 ∗ PCI (roughness)

(7)

OSCI = 1 ∗ SCI (deflection)

(8)

5.8 Evaluation of Overall Pavement Condition Index (UR-OPCI) The Overall Pavement Condition Index (UR-OPCI) was determined based on (OFCI) and (OSCI). The weightage was individually applied with structural and functional pavement criteria to define the final overall pavement condition index for best results. The weight values were calculated by Expert Choice 11 software. From the evaluation of 125 surveys, functional factors received a weighting of 60% and, as a result, structural parameters 40%; the final UR-OPCI was calculated using Eq. 9. The calculated value of PCI (distress), PCI (roughness), PCI (deflection), OFCI, OSCI, and OPCI are listed in Table 3. OPCI = 0.6 ∗ (OFCI) + 0.4 ∗ (OSCI)

(9)

12

S. Singh Yadav et al.

Table 3 UR-OPCI value ranges UR-OPCI range

Pavement rating

Maintenance and repair strategies

0–15

Very poor

Full-depth reconstruction, reclaimed asphalt pavement recycling

15–30

Poor

Thick overlays, premix, carpet, surface dressing

30–50

Fair

Thick overlays, full-depth patching, pothole filling

50–65

Good

Thin overlays, patching, fog seal

65–80

Very good

Thin overlays, chip seal, micro-surfacing

80–100

Excellent

Routine maintenance that includes micro-crack sealing patching

5.9 UR-OPCI-Based Maintenance and Repair Strategies The current study proposes specific maintenance and repair techniques for preventive and corrective actions in urban road sections based on different value ranges of the Overall Pavement Condition Index (UR-OPCI). Because the Overall Pavement Condition Index (UR-OPCI) was developed using the functional and structural parameters of urban roads, it is considered the best indicator of road surface condition. Table 3 recommends maintenance and repair procedures according to different UR-OPCI value ranges [9].

6 Results and Discussion a. Figure 5 shows individual functional condition indices and structural condition indices. Road R13 shows the highest PCI (distress) value and R24, R27, R39 show the lowest PCI (distress) value. R1 shows the highest PCI (roughness) value and R47 shows the lowest PCI (roughness). Similarly, R4 shows the highest SCI (deflection) and R7 shows the lowest SCI (deflection) (Table 4).

PCI & SCI(0-100)

Individual Functional and Structural Condition Index 120 80 40 0

PCI(Distress)

PCI(IRI)

SCI(D)

R1 R3 R5 R7 R9 R11 R13 R15 R17 R19 R21 R23 R25 R27 R29 R31 R33 R35 R37 R39 R41 R43 R45 R47

Section ID

Fig. 5 Individual functional and structural indices

Development of Maintenance Priority Index for Urban Road Network

13

Table 4 Calculated value of PCI (Distress), PCI (IRI), SCI (D), OFCI, OSCI and UR-OPCI, and ranking based on UR-OPCI for selected urban road network ID

PCI (ASTM)

PCI (distress)

PCI (IRI)

SCI (D)

OFCI

OSCI

OPCI

Ranking based on UR-OPCI

R1

98

45.51

66.29

29.9

52.78

29.9

43.63

26

R2

100

39.15

66.03

19.1

48.56

19.1

36.77

15

R3

68

99.70

61.48

40.7

86.32

40.7

68.07

47

R4

85

52.34

56.05

73.9

53.64

73.9

61.74

46

R5

79

83.41

56.03

24.5

73.83

24.5

54.10

42

R6

79

51.35

53.27

34.3

52.02

34.3

44.93

27

R7

72

83.18

52.56

6.33

72.46

6.33

46.01

30

R8

97

85.95

50.16

13.6

73.42

13.6

49.49

36

R9

89

44.97

48.85

15.4

46.33

15.4

33.96

8 16

R10

100

48.34

48.11

21.8

48.26

21.8

37.68

R11

100

39.16

47.43

18.2

42.05

18.2

32.51

5

R12

78

81.67

46.91

16.8

69.51

16.8

48.42

34 48

R13

79

99.93

46.50

57.5

81.23

57.5

71.74

R14

100

32.29

45.88

17.1

37.05

17.1

29.07

2

R15

100

51.34

45.67

14.7

49.36

14.7

35.50

11

R16

100

41.22

45.16

14.3

42.59

14.3

31.28

4

R17

83

49.40

45.14

31.9

47.91

31.9

41.51

21

R18

88

60.99

45.09

22.1

55.43

22.1

42.10

22

R19

89

39.16

44.28

30.1

40.95

30.1

36.61

13

R20

72

67.52

43.97

50.7

59.28

50.7

55.85

43

R21

100

48.34

42.97

50.7

46.46

50.7

48.15

33

R22

100

32.29

42.69

42.7

35.93

42.7

38.64

18

R23

100

83.39

42.47

40.3

69.07

40.3

57.56

44

R24

100

31.33

42.45

65.3

35.22

65.3

47.25

31

R25

100

64.18

42.39

34.8

56.52

34.8

47.83

32

R26

100

39.15

42.30

34.8

40.16

34.8

38.02

17

R27

100

31.33

42.04

19.6

35.01

19.6

28.85

1

R28

100

51.34

41.86

56.5

47.79

56.5

51.27

39

R29

100

60.97

41.19

25.1

54.03

25.1

42.46

23

R30

78

94.95

41.13

16.8

76.06

16.8

52.36

40

R31

100

96.96

40.99

9.91

77.31

9.91

50.35

38

R32

96

77.55

40.82

24.3

64.58

24.3

48.47

35 20

R33

100

66.73

40.51

13

57.43

13

39.66

R34

100

54.04

40.16

11.1

49.11

11.1

33.90

7

R35

62

92.13

39.94

40.5

73.81

40.5

60.49

45 (continued)

14

S. Singh Yadav et al.

Table 4 (continued) ID

PCI (ASTM)

PCI (distress)

PCI (IRI)

SCI (D)

OFCI

OSCI

OPCI

Ranking based on UR-OPCI

R36

89

50.88

39.79

18.5

46.92

18.5

35.55

12

R37

76

83.38

39.58

22.4

67.88

22.4

49.69

37

R38

87

50.88

39.10

37.5

46.63

37.5

42.98

24

R39

93

31.33

38.74

31.5

33.72

31.5

32.83

6 29

R40

73

83.24

38.66

13.5

67.35

13.5

45.81

R41

100

41.22

38.16

15.1

39.95

15.1

30.01

3

R42

62

67.52

37.85

23.2

56.79

23.2

43.35

25 28

R43

68

67.53

37.59

28

56.63

28

45.18

R44

70

54.04

36.86

13.9

47.83

13.9

34.26

9

R45

67

64.18

36.39

15.6

54.39

15.6

38.87

19

R46

68

51.35

36.12

18.3

46.02

18.3

34.93

10

R47

88

50.86

36.05

23.3

45.67

23.3

36.72

14

R48

69

83.35

38.27

32.3

67.57

32.3

53.46

41

b. Figure 6 Comparison of PCI calculated using the ASTM method and overall pavement condition index (UR-OPCI) using the methodology used in the present study. R13 shows the highest UR- OPCI and R27 shows the lowest UR-OPCI value. c. UR-OPCI can also prove to be cost-effective and give correct strategic maintenance priority of urban road network. As a result of the above discussion, it is clear that the Overall Pavement Condition Index (UR-OPCI) is an accurate tool for identifying the priority ranking for maintenance plans of various urban road sections. Comparison B/W PCI(ASTM) &UR- OPCI

PCI & UR-OPCI

PCI(ASTM)

OPCI

150 100 50 0

R1 R3 R5 R7 R9 R11 R13 R15 R17 R19 R21 R23 R25 R27 R29 R31 R33 R35 R37 R39 R41 R43 R45 R47

Section ID

Fig. 6 Comparison B/W PCI (ASTM) and UR-OPCI

Development of Maintenance Priority Index for Urban Road Network

15

7 Conclusions The current study examined urban roads to prioritize the urban road network and effective use of road maintenance funds. The current study assessed pavement maintenance through functional and structural assessments of urban roads. 48 urban road segments in DELHI-NCR were selected for the development of Roughness Index model, a Deflection model, a Distress model and finally an Overall pavement Condition Index (UR-OPCI). The following conclusions can be drawn from this study— a. The objectives were achieved after a detailed collection of the data on structural and functional assessment aspects, which aided in the creation of roughness index models, distress models, and pavement surface deflection models, all of which led to the formulation of OPCI. b. According to study of the Overall Pavement Condition Index (UR-OPCI), R13 has UR-OPCI in the range of 65–80, indicating that pavement is in very good condition, it requires Thin Overlays, Chip Seal, and Micro-surfacing. c. According to UR-OPCI ranking, R27 has the highest priority and should be serviced first, while R13 is the best road and should be serviced last. d. The UR-OPCI leads to an effective allocation of maintenance resources and is proving to be a valuable tool for road maintenance engineers and traffic managers. It also contributes to the long-term growth of the country.

Annexure—Format of questionnaire survey Format of Questionnaire survey for functional and structural parameters of pavement

Somewhat more important

Much more important

Very much more important

Absolutely more important

Intermediate value when compromise is required between above

5

7

9

2,4,6,8

9

8

7

6

5

4

3

2

1

2

3

4

9

8

7

6

5

4

3

2

1

2

Group 1 parameter

9

8

7

6

5

4

3

2

1

2

For example: If you think that road rutting has equal importance to road ravelling then put (

Rutting

Group 1 parameter

3

3

6

4

7

8

9

Ravelling

Group 2 parameter

5

6

5

6

7

7

8

8

9

9

(continued)

Group 2 parameter

Ravelling

Group 2 parameter

) under 5 on right side towards ravelling

5

) under 5 on left side towards cracking

) under 1 in the centre

4

For example: If you think that road rutting is 5 times more important than road cracking then put (

Rutting

Group 1 parameter

For example: If you think that road rutting is 5 times more important than road ravelling then put (

Sample

Please check the box with a tick (

) for the relative importance between group 1 and group 2 parameters

Equal importance

1

3

Part-A

Definition

Intensity of importance

Fundamental scale

Relative importance of parameters corresponding to pavement performance and maintenance

Questionnaire

16 S. Singh Yadav et al.

Rutting

Intensity of importance

Definition

Fundamental scale

Relative importance of parameters corresponding to pavement performance and maintenance

Questionnaire

(continued)

Ravelling

Development of Maintenance Priority Index for Urban Road Network 17

18

S. Singh Yadav et al.

References 1. Boyapati B, Kumar RP (2015) Prioritisation of pavement maintenance based on pavement condition index. Indian J Sci Technol 8. doi:https://doi.org/10.17485/ijst/2015/v8i14/64320 2. Al-Neami M, Al-Rubaee R, Kareem Z (2018) Assessment of Al-Amarah street within the Al-kut city using pavement condition index (PCI) and GIS technique. In: MATEC web of conferences, vol 162. EDP Sciences, p 01033 3. Murshan M, Vala M (2017) Developing methodology for priority of pavement maintenance for urban arterial. Int J Adv Eng Res Dev (IJAERD) 4(4). doi:https://doi.org/10.21090/ijaerd. 19035 4. Tariq M, Pimplikar SS (2017) A comparative study on pavement condition rating methods for flexible roads. Int J Eng Dev Res 5:1255–1260 5. Janani L, Dixit RK, Sunitha V, Mathew S (2020) Prioritisation of pavement maintenance sections deploying functional characteristics of pavements. Int J Pavement Eng 21(14):1815– 1822 6. Zhang Z, Manuel L, Damnjanovic I, Li Z (2003) Development of a new methodology for characterizing pavement structural condition for network-level applications. Texas Department of Transportation, Austin, TX 7. Solanki U (2014) A review on structural evaluation of flexible pavements using falling weight deflectometer. STM J 2:1–10 8. Shah Yogesh U, Jain SS, Tiwari D, Jain MK (2013) Development of overall pavement condition index for urban road network. Procedia Soc Behav Sci 104:332–341, ISSN 1877-0428. doi:https://doi.org/10.1016/j.sbspro.2013.11.126 9. Gupta A, Gupta A; Kumar P, Kumar A (2021) Rural road maintenance prioritization index based on functional and structural parameters for rural road network in Himachal Pradesh 10. Shahnazari H, Tutunchian M, Mashayekhi M, Amini AA (2012) Application of soft computing for prediction of pavement condition index. J Transp Eng 138:1495–1506. doi:https://doi.org/ 10.1061/(ASCE)TE.1943-5436.0000454 11. Zhang Z, Murphy MR, Peddibhotla S (2011) Implementation study of a structural condition index at the network level. In: Eighth international conference on managing pavement assets, ISBN: 9789561412309, TRID 12. Vyas V, Singh AP, Srivastava A (2021) Prediction of asphalt pavement condition using FWD deflection basin parameters and artificial neural. Road Mater Pavement Des 22(12):2748–2766, Taylor & Francis. doi:https://doi.org/10.1080/14680629.2020.1797855 13. Arhin SA, Williams LN, Ribbiso A, Anderson MF (2015) Predicting pavement condition index using international roughness index in a dense urban area. J Civ Eng Res 5(1):10–17, p-ISSN: 2163-2316, e-ISSN: 2163-2340. doi:https://doi.org/10.5923/j.jce.20150501.02 14. Al-Neami MA, Al-Rubaee RH, Kareem ZJ (2017) Evaluation of pavement condition index for roads of Al-Kut city. Int J Curr Eng Technol 7(4)

Modelling of Deflection Basin Parameters of Asphalt Pavements Using Artificial Neural Networks and Adaptive Neuro-Fuzzy Inference Systems Sachin Gowda, K. Vaishakh, Aakash Gupta, R. Prakash, and G. Kavitha

Abstract Non-destructive testing equipment, such as the Falling Weight Deflectometer, offers crucial evaluations of the structural state of the road and enhances pavement management systems. Various approaches based on pavement surface deflection measured using Falling weight deflectometers are widely used around the world for assessing structural stability. The backcalculation of pavement layer moduli has been a widely recognized approach for assessing the structural adequacy of the pavement. However, consistently performing these tests at the network level is laborious, and the subsequent interpretation of the data requires technical expertise, a great deal of time, finance, and other resources. Because of this structural component of roadways, decisions when choosing between maintenance and repair are often neglected. This study uses a variety of structural, functional, environmental, and subgrade soil properties as input parameters to develop a trusted relationship for the estimation of seven different deflection basin parameters such as surface curvature index, Base Curvature Index, Base Damage Index, Area Under Pavement Profile, Deflection Ratio, Shape factors F1 and F2. An effective model was developed using artificial intelligence-based soft computing techniques; Artificial Neural Networks (ANN) and Adaptive Neuro-fuzzy Inference Systems (ANFIS) to predict the output deflection basin parameters from the input variables. The data to train, test and validate the model were gathered through field trials. To achieve the above goal, several models based on ANN and ANFIS were trained by changing number of hidden layers, the neurons in the layer and number of membership functions. Prediction efficiency of the model is assessed based on its root mean square error and the coefficient of determination value.

S. Gowda (B) · A. Gupta CSIR-Central Road Research Institute (CRRI), New Delhi 110025, India e-mail: [email protected] A. Gupta e-mail: [email protected] K. Vaishakh · R. Prakash · G. Kavitha RASTA-Center for Road Technology, Bengaluru 560058, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_2

19

20

S. Gowda et al.

Keywords Surface curvature index · Base curvature index · Base damage index · Area under pavement profile · Deflection ratio · Shape factors

1 Introduction Pavement surface deflection has been utilized by a number of agencies to measure its structural performance. To measure the deflection of the road surface under loading, devices like deflectograph and falling weight deflectometers are frequently employed. In project-level studies, structural assessments and remediation solutions are often provided using mechanistic approaches and back-calculation techniques. Decisions concerning the choice and use of appropriate procedures for pavement restoration are frequently made using data on the functional and structural state of the pavement. Many highway agencies have utilized the La Croix deflectograph and Benkelman beam extensively to measure roadway deflection and determine the need for repair and reconstruction. According to Kim et al. [1]; Talvik [2], the surface curvature index (SCI), more closely associated to the properties of bound layer, is the difference between the deflection at the centre and the deflection at 300 mm radial offset from the centre (D0 –D300 ). The parameter (D300 –D600 ) is called the as the Base Damage Index (BDI) and is more closely related to the base layer, while (D600 –D900 ) represents the Base Curvature Index (BCI) and reflects the condition of the base layer and subgrade. Area Under Pavement Profile (AUPP), another property derived from deflection, has been used effectively to quantify pavement stiffness as well as to link the horizontal strain at the base of bituminous layer. The shape parameters F1 and F2 help determine the state of the layer at a given depth by representing the degree of curvature of the deflection basin. As a result, various rapid analysis techniques are required to quantify the structural health of the pavement using FWD data. Utilizing DBPs is one of the approaches that aids in determining the geometry of the deflection basin under the load. The usefulness of DBP in assessing the structural condition of roads has been highlighted by a number of researchers in the past [1–8]. The commonly proposed DBP includes the base damage index, the SCI, the BCI, the AUPP [1, 6–12]. The area ratio as well as normalized area ratio were used effectively to check the structural health of pavement [13]. In order to draw conclusions on the structural health of pavement layers, this study uses Deflection Basin Parameters (DBPs). Two different AI-harnessed models were developed to indirectly measure the DBP from input variables that are relatively easily collectable. The results of the study are presented in this paper, which supports the use of AI models for the interpretation of pavement condition.

Modelling of Deflection Basin Parameters of Asphalt Pavements Using …

21

1.1 Objectives This primary goal of the research was to evaluate the usefulness of Deflection Basin Parameters (DBPs) generated during FWD testing as network-level indicators of highway structural health. To assess the functional condition of the pavement, distress survey was performed using Network Survey Vehicle (NSV). The output variables in this study are the most commonly used DBPs such as SCI, BCI, BDI, AUPP, DR and Shape Factors F1 and F2, while the input variables are pavement structure-related variables such as Bituminous layer thickness (La), Granular layer thickness (Lb) and; functional performance variables such as the % area of Cracking; and Subgrade soil strength indicators such as Soil type, Plasticity Index, Maximum Dry Density (MDD) and California Bearing Ratio (CBR); and finally environment variables such as pavement temperature and ambient temperature. The choice of input variables is primarily influenced by the accessibility and ease of collecting these data compared to that of FWD. Contrarily, as the entire method— from data collection to back-calculation analysis—must be followed each time with FWD testing, the current situation demands a method that reduces the frequency of FWD testing. This study emphasizes on combining variables from several areas to produce accurate and highly flexible prediction models.

2 Analysis To find a suitable way to evaluate the correlation between the DBPs and Cracking, PI, MDD, CBR, Bituminous layer thickness, Granular layer thickness, Air temperature and Pavement temperature ANN, and ANFIS analysis were carried out and compared. The associated methods are briefly outlined below.

2.1 Artificial Neural Network ANN is an analogue of a minimalistic illustration of how the biological human brain is put together. Recently, academics working on advanced technical problems has used ANN to solve these challenges. Layers of interconnected neurons—the basic processing unit constitutes the conventional ANN architecture. The many connections that exist between these neurons and their ability to learn from their input allow the output to be predicted effectively. A set of input nodes, some hidden nodes at one or more levels, and some output nodes form the basic structure of the ANN. The training function, the transfer function, the number of hidden layers, and the number of neurons in those layers all affect how well the model can predict the output. By combining multi-layer perceptron, feed-forward-back propagation, logsig (log-sigmoid) transfer function and Levenberg–Marquardt learning function, a

22

S. Gowda et al.

Fig. 1 Architecture of ANN model

suitable ANN model to predict the output DBPs was developed (trainlm). Several combinations of the number of hidden layers and the right number of neurons were examined to get a model with a better R2 value and a lesser RMS error. The ANN model was developed with single hidden layer for nine inputs: cracking, PI, MDD, CBR, bituminous layer thickness, granular layer thickness, air temperature and pavement temperature and an output deflection basin parameter such as SCI, BCI, BDI, deflection ratio, area under pavement profile, shape factor F1 and shape factor F2. Of the total data points, 70% were used to train the model. 15% were used to test the model and the remaining 15% of data were validated using MATLAB r2022a software. The design of the proposed ANN model is shown in Fig. 1.

2.2 Adaptive Neuro-Fuzzy Inference System Using a hybrid algorithm, the ANFIS model was evolved that incorporated both ANN and fuzzy logic with a series of fuzzy language rules, producing Input-output models resembling human-like knowledge and specific input–output data combinations. Membership functions in ANFIS modelling signify how fuzzy the data set is. Based on its geometry, they are classified as triangular, bell-shaped, trapezoidal and Gaussian membership functions. The trapezoidal membership function (Trapmf) with different epoch numbers was used to identify the model with the lowest RMSE value for the ANFIS study. Takagi–Sugeno type ANFIS with Grid partitioning was employed in the construction of FIS. In addition, the resulting FIS was trained using 70% of the data using a hybrid learning algorithm. The model was validated on 15% of the data and was tested on the remaining 15% of the data using MATLAB r2022a. The structure of the proposed model is illustrated in Fig. 2.

Modelling of Deflection Basin Parameters of Asphalt Pavements Using …

23

Fig. 2 Proposed architecture of ANFIS model

2.3 Performance Criteria The potential of the ANN and ANFIS models was assessed statistically by computing the RMSE, R2 values. These criteria are defined by the equations below. /E RMSE =

n j=1

(E − A)2 n

E = Predicted Value A = Observed Value E ( A i − E i )2 R2 = 1 − E (Ai − Ai )2 A = Average of A values n = Number of Observation

24

S. Gowda et al.

3 Results and Discussions 3.1 ANN Results The ANN model was developed with single hidden layer for nine inputs: cracking, PI, MDD, CBR, bituminous layer thickness, granular layer thickness, air temperature and pavement temperature and an output deflection basin parameter such as SCI, BCI, BDI, deflection ratio, area under pavement profile, shape factor F1 and shape factor F2. To create an effective model with the lowest RMSE and highest R2 value, the architecture of the models was varied by altering the number of neurons in the hidden layer. A logsig transfer function was applied to this hidden layer of different number of neurons. A total of seven different models were developed for different outputs with same inputs. The test of the model for predicting the output was established through training and validation data. Model performance of training, validation, testing for RMSE and R2 are given, respectively, the following Table 1 gives the results and in Fig. 3, the R2 values obtained for the network structure SCI (9-10-1).

3.2 ANFIS Results The ANFIS model was developed using two trapezoidal membership functions (trapmf) for each input variable to determine the deflection basin parameters. The input parameters such as cracking, PI, MDD, CBR, bituminous layer thickness, granular layer thickness, air temperature and pavement temperature and different deflection basin parameters such as SCI, BCI, BDI, deflection ratio, area under pavement profile, shape factor F1 and shape factor F2 as output resulted in seven different models. Table 2 summarizes their performance in terms of RMSE and R2 value and Fig. 4, shows the regression results for SCI using ANFIS.

4 Discussion The artificial intelligence harnessed soft computing techniques; ANN and ANFIS used in this work have successfully shown how input and output variables correlate with each other. The R2 values in the range of 0.66–0.93 for the ANN model and 0.45–0.86 for ANFIS model depict of strong correlation between the input and output parameters. The RMSE error for the models ranged from 0.014–0.335 to 0.007– 0.416, respectively. Although it is observed that the R2 values are satisfactory, it should be highlighted that the data used for modelling came from meticulous field testing and might as well involve human errors in the data collection. The robustness of these models is ensured by the big dataset and successful field testing, and their application has useful field consequences. Even though this research limited the

Modelling of Deflection Basin Parameters of Asphalt Pavements Using …

25

Table 1 Test results of ANN model Architecture

Neurons in hidden layer

Output parameters

9-10-1

10

SCI (mm)

9-14-1

9-11-1

9-14-1

9-13-1

9-15-1

9-14-1

14

11

14

13

15

14

BCI (mm)

BDI (mm)

Deflection ratio

AUPP (mm)

Shape factor F1

Shape factor F2

Model performance

RMSE

R2

Training

0.014

0.927

Validation

0.016

0.912

Testing

0.015

0.901

Training

0.008

0.906

Validation

0.009

0.872

Testing

0.009

0.900

Training

0.019

0.837

Validation

0.021

0.815

Testing

0.020

0.776

Training

0.022

0.846

Validation

0.023

0.787

Testing

0.025

0.812

Training

0.079

0.815

Validation

0.092

0.770

Testing

0.096

0.762

Training

0.072

0.779

Validation

0.075

0.710

Testing

0.082

0.754

Training

0.301

0.708

Validation

0.305

0.720

Testing

0.335

0.658

number of hidden layers to unity for the convenience of modelling for a very big database, the R2 values can still be enhanced in the case of ANN models by adjusting the number of hidden layers. When dealing with ANFIS, experimenting with various combinations of membership functions and increasing the number of membership functions can lead to a higher R2 value. However, for two membership functions, this research employed trapezoidal membership function, which provided higher coefficient of correlation and least RMSE value compared to other membership functions. The research successfully demonstrates the ease and capability of artificial intelligence-based soft computing techniques to surpass complex challenges of modelling pavement reactions, where a variety of diverse components play important roles. By doing such comparable, reliable models would help respective authorities speed up the decision-making processes for pavement maintenance and rehabilitation measures.

26

S. Gowda et al.

Fig. 3 Regression results obtained for the network structure SCI (9-10-1)

In general, the value of the coefficient of determination, which measures how well the input and output parameters are related, is found to be high (close to one), and the mean square error, which is the mean squared difference between the actual outputs and predicted output, is found to be equally low with reasonable accuracy. However, appropriate numbers would vary from situation to situation and depend on the availability of data. The average R2 value of the ANN models developed in the study is 0.81 throughout training, validation and testing. The highest R2 value was observed in predicting SCI, and the lowest value was observed in predicting shape factor F2 (Table 1). Similarly, the average R2 value of the ANFIS models developed in the study throughout training, validation and testing is 0.64. Also, in this model, the highest R2 value was observed in predicting SCI and the lowest value in predicting shape factor F2 (Table 1). The higher R2 values demonstrate the significance of the correlation generated for the seven models developed using ANN and ANFIS, respectively, in this study, which is further supported by the low MSE values.

Modelling of Deflection Basin Parameters of Asphalt Pavements Using …

27

Table 2 Test results of ANFIS model Number of membership functions

Output parameters

2

SCI (mm)

2

2

2

2

2

2

BCI (mm)

BDI (mm)

Deflection ratio

AUPP (mm)

Shape factor F1

Shape factor F2

Model performance

RMSE

R2

Training

0.017

0.848

Validation

0.026

0.733

Testing

0.031

0.662

Training

0.007

0.858

Validation

0.012

0.662

Testing

0.017

0.675

Training

0.013

0.792

Validation

0.020

0.671

Testing

0.021

0.675

Training

0.022

0.721

Validation

0.031

0.490

Testing

0.037

0.467

Training

0.076

0.726

Validation

0.102

0.621

Testing

0.138

0.612

Training

0.067

0.673

Validation

0.089

0.582

Testing

0.118

0.541

Training

0.284

0.586

Validation

0.387

0.432

Testing

0.416

0.446

5 Conclusions The appropriateness of utilizing AI-powered models for the prediction of structural performance parameters in asphalt pavements is justified in this research. The DBPs, namely SCI, BCI, BDI, AUPP, DR, F1 and F2, have been modelled using ANN and ANFIS to indirectly measure it from the structural, functional, environmental and subgrade soil properties. The study developed seven models for each of the DBPs as outputs using both ANN and ANFIS approach. The dataset to train, validate and test the model was obtained from field studies conducted by performing a FWD and NSV study. A big dataset of 2001 records were gathered via field testing. Further laboratory tests were performed to assess the subgrade soil properties. Seven ANN structures as 9-10-1, 9-14-1, 9-11-1, 9-14-1, 9-13-1, 9-15-1 and 9-14-1 for each of the seven output parameters independently to achieve more accurate modelling perspectives. The ANN model predicted better with R2 value as high as 0.93 and RMSE value as low as 0.014. Similarly, seven ANFIS models were generated for

28

S. Gowda et al.

Fig. 4 Regression results for SCI using ANFIS

each of the seven output DBPs. The model could provide a prediction efficiency with a R2 value as high as 0.086 and RMSE value as low as 0.007. The higher R2 values in predicting the Surface Curvature Index explain why the asphalt layer’s characteristics have a greater overall influence on pavement quality. The results demonstrate the higher prediction efficacy of ANN model compared to ANFIS model. The study’s preliminary methodology offers valid relationship between the adopted input parameters to the Deflection bowl parameters. This quick and straightforward method of data analysis reduces the requirement for the laborious backcalculation procedure. In addition to providing a comprehensive knowledge of the pavement layers that contribute to the current state, the proposed models would aid in estimating Deflection bowl parameter values, which directly indicate of the structural condition of the pavement. This will substantially reduce the cumbersome data collection through FWD due to the simplicity of gathering input dataset. This will, however, reduce the dependency of Maintenance and Rehabilitation agency on the functional condition of the pavement.

Modelling of Deflection Basin Parameters of Asphalt Pavements Using …

29

References 1. Kim YR (2000) Assessing pavement layer condition using deflection data (Report No. 10-48). National Research Council, Transportation Research Board (US) 2. Talvik O, Aavik A (2009) Use of FWD deflection basin parameters (SCI, BDI, BCI) for pavement condition assessment. Baltic J Road Bridge Eng 4(4):196–202 3. Carvalho R, et al (2012) Simplified techniques for evaluation and interpretation of pavement deflections for network-level analysis. Office of Infrastructure Research and Development, Federal Highway Administration, Report No. FHWA-HRT-12-023 4. Donovan PR (2009) Analysis of unbound aggregate layer deformation behaviour from full scale aircraft gear loading with wander. Doctoral dissertation, University of Illinois at UrbanaChampaign, Urbana, Illinois 5. Gopalakrishnan K, Thompson MR (2005) Use of deflection basin parameters to characterize structural degradation of airport flexible pavements. In: Proceedings of Geo-Frontiers Congress. pp 1–15 6. Horak E (1987) Aspects of deflection basin parameters used in a mechanistic rehabilitation design procedure for flexible pavements in South Africa [Unpublished doctoral dissertation]. https://repository.up.ac.za/handle/2263/23960 7. Horak E (2008) Benchmarking the structural condition of flexible pavements with deflection bowl parameters. J S Afr Inst Civ Eng 50(2):2–9 8. Horak E, Emery S, Maina J (2015) Review of falling weight deflectometer deflection benchmark analysis on roads and airfields. In: Conference on asphalt pavement for South Africa (CAPSA2015), Sun City, South Africa 9. Losa M, Bacci R, Leandri P (2008) A statistical model for prediction of critical strains in pavements from deflection measurements. Road Mater Pavement Des 9(sup1): 373–396. doi:https:// doi.org/10.1080/14680629.2008.9690175 10. Park HM, Kim YR, Wan Park S (2005) Assessment of pavement layer condition with use of multiload-level falling weight deflectometer deflections. Transp Res Rec: J Transp Res Board 1905(1):107–116. https://doi.org/10.1177/0361198105190500112 11. Xu B, Ranjithan SR, Kim YR (2002) New condition assessment procedure for asphalt pavement layers, using falling weight deflectometer deflections. Transp Res Rec: J Transp Res Board 1806(1):57–69. https://doi.org/10.3141/1806-07 12. Xu B, Ranjithan SR, Kim YR (2002) New relationships between falling weight deflectometer deflections and asphalt pavement layer condition indicators. Transp Res Rec: J Transp Res Board 1806(1):48–56 13. Saleh M (2016) Simplified approach for structural capacity evaluation of flexible pavements at the network level. Int J Pavement Eng 17(5):440–448. https://doi.org/10.1080/10298436.2014. 993202

Comparative Studies on Gel-Incorporated Flexible Pavement Delvin J. Joseph, Padmakumar Radhakrishnan, and Vignesh Dhurai

Abstract The climatic changes we are currently experiencing are so drastic that the existing pavements are not able to keep up with them; among these, the floods are more severe as they occur more frequently than other hazards. Aerogel is a substance that is highly resistant to water. This paper presents the study of application of aerogel in bituminous pavement construction. Since silica-gel is having similar features, for comparison silica-gel modified pavement is created, which is used to compare the results obtained with the aerogel-incorporated bitumen pavement. The FEM analysis results of both silica-gel and aerogel-modified bitumen are compared with the standard bitumen pavement analysis results. Similarly, aerogel-incorporated concrete pavement model is created, and results are compared with the standard concrete pavement. The aerogel-incorporated pavement performs well compared to silica gel while maintaining the required strength required for the standard pavement. Keywords Aerogel · Finite element method · Water resistant pavement · Gelincorporated pavement

1 Introduction The primary objective of a pavement structure is to provide a surface that offers acceptable riding quality, skid resistance, light-reflecting characteristics, and low noise pollution. The aim is to ensure that the stresses transmitted due to wheel loads are reduced sufficiently, so they do not exceed the bearing capacity of the sub-grade. D. J. Joseph (B) · P. Radhakrishnan · V. Dhurai Department of Civil Engineering, College of Engineering Trivandrum, Thiruvananthapuram, Kerala, India e-mail: [email protected] P. Radhakrishnan e-mail: [email protected] V. Dhurai e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_3

31

32

D. J. Joseph et al.

There are two main types of pavements that serve the purpose: flexible pavement and rigid pavement. An ideal pavement must meet certain requirements to perform its functions effectively. These requirements include having sufficient thickness to distribute wheel load stresses to a safe value on the subgrade soil, being structurally strong enough to withstand all types of stresses imposed upon it, having an adequate coefficient of friction to prevent skidding of vehicles, having a smooth surface to provide comfort to road users even at high speed, producing minimal noise from moving vehicles, having a dust-proof surface that does not reduce visibility and an impervious surface that protects the sub-grade soil. Furthermore, an ideal pavement should have a long design life and low maintenance cost. Apart from these requirements, one of the most critical aspects that an ideal pavement must meet is the ability to withstand the detrimental effects of water over its lifetime. Thus, it can be summarized that a pavement structure should be designed to meet all these requirements to provide safe and comfortable driving conditions. Silica aerogel particles are used to prepare lightweight concrete by replacing normal aggregates of concrete, the resulting mix was lightweight and thermal insulating concrete material, and the aerogel particles were stable during the hydration of cementitious materials [2]. The application of aerogel and their potential in heritage buildings along with technical properties of commercially available aerogel materials, super-insulating aerogel materials have an exceptional potential in the refurbishment of heritage buildings [3]. The aerogel can manufacture with locally available materials while large specific surface area, high extent of porosity and very low density followed by a microstructure in the form of interconnected pores and channels making these materials very fascinating for most of their high-performance applications [4]. Types, properties, industrial applications, manufacturing procedures and products of aerogel, modified products like oxide aerogels, polymeric aerogels, mixed aerogels, hybrids, composites with fibres etc. help to use them in different fields due to their wide varieties of properties [5]. Silica gel is a compound capable of absorbing water to their structures, asphalt mix at different combination of silica gels was compared with hot mix and warm mix asphalt with zeolites, and the results were like that with zeolites [6]. From the literature review, no study was reported on the evaluation of the potential of gels like aerogel in enhancing the moisture resistance and thereby the durability of pavements. The aim of this study is to compare the performance of gel-incorporated pavement with hot-mix asphalt pavement using the aid of finite element analysis by ABAQUS software.

Comparative Studies on Gel-Incorporated Flexible Pavement

33

2 Methodology 2.1 Aerogel Aerogel, being a nanostructured, open porous solid made via sol–gel technology was selected for the pavement construction. There are several types of aerogels, some of them are: . . . .

Oxides (Quarz, Titania, Zirkonia. Mixed Oxides) Polymers (Resorcin-, Melamin-Formaldehyde) Carbon (Pyrolyzed Polymere) Cellulose, starch etc. and almost everything that can be gelled (Table 1).

Among the above types of aerogels, silica-based aerogel is selected as they are available in abundance and easy to produce (Table 2).

2.2 Pavement Model Using ABAQUS™ A three-dimensional finite element model used for a typical flexible pavement section designed in ABAQUS™ software consists of bitumen surface layer, a base layer, a Table 1 Main properties and typical values of aerogels depending on their composition [4] Composition

Bulk density (gcm−3 )

Specific surface area ( m2 g−1 )

Silica aerogel

0.003–0.5

600

Resorcinol formaldehyde aerogel

0.005–0.3

100–1500

Carbon aerogel

0.05

100–1000

Cellulose aerogel

0.1–0.35

200–400

Table 2 Main physical properties of SiO2 aerogels [4] Range

Properties Bulk density

(kg/m3 )

Skeletal density (g/cm3 )

3–150

Typical value 100

1.7–2.1

Porosity (%)

90–99.8

Mean pore diameter (nm)

20–150

Inner surface area (m2 /g)

500–1500

1000

Refractive index

1.007–1.24

1.02

Thermal conductivity λ (in air 300 K) (Wm2 /g)

0.014–0.021

0.015

Young’s modulus E (Mpa)

0.002–100

1

Sound velocity (m/s)

20–800

100

34

D. J. Joseph et al.

Fig. 1 Finite element model of standard pavement

subbase layer and a subgrade layer, which are all simulated by 3D deformable solid extrusion elements (solid homogeneous section type). The pavement surface was designed about 880 × 880 mm dimension and having a total thickness of 880 mm. The depth 880 mm is subdivided into four layers with thickness, respectively, as surface layer (80 mm), base layer (100 mm), subbase layer (200 mm) and subgrade (500 mm) (Fig. 1 and Table 3).

2.3 Load and Boundary Conditions The design of pavement systems is heavily influenced by traffic load conditions, which include factors such as axial loads, axle configurations, tire contact areas, number of load repetitions and vehicle speed. Heavy vehicular traffic, particularly from trucks, is a significant cause of pavement distress and failure. In numerical and theoretical analyses, the most common method of applying wheel load is through uniformly distributed tire pressure loads on a circular equivalent contact area based on the wheel load and tire contact pressure. The wheel load is equal to half of the axle load and can be applied to the entire wheel path in the case of lading areas. To simulate heavy traffic, an impulse-type load was used with an amplitude equal to half of an axle load (PAxle = 80 kN, so P = 40 kN), resulting in a pavement stress of 0.56 MPa. This surface wheel load (P) was applied to a circular pressure contact area with a 150 mm radius at the centre of model domain. Boundary conditions were set for the model, with the bottom of the model constrained and displacements and rotations prevented on the sides parallel to the x and y axes. To accurately estimate the stress field in the pavement road section, the degree of mesh refinement is critical, with the

Comparative Studies on Gel-Incorporated Flexible Pavement

35

Table 3 Properties of each layer and substances [1, 3, 6] Layer Surface

Thickness (mm)

Density × 10–9 (tonne/mm3 )

Young’s modulus (Mpa)

Poisson’s ratio

80

2.332

2,000

Base

100

2.162

300

0.35

0.3

Subbase

200

1.922

600

0.35

Subgrade

500

1.762

76

0.45

Silica-gel bituminous pavement (SGBP)

80

2.859

1,000

0.35

Concrete pavement

80

2.4

40,000

0.15

Aerogel incorporated concrete pavement (AICP)

80

1

5,533

0.2

Aerogel incorporated bituminous pavement (AIBP)

80

0.2

659

0.2

densest mesh required for the superficial layer where the load is applied to capture permanent displacements. The pavement layers were meshed using eight-node linear brick elements (C3D8R). Since traffic load conditions have a significant impact on the design of pavement systems, with heavy vehicular traffic being a primary cause of pavement distress and failure. The most common method of applying wheel load in numerical and theoretical analyses is through uniformly distributed tire pressure loads on a circular equivalent contact area. To accurately estimate the stress field in the pavement road section, the degree of mesh refinement is crucial, with a denser mesh required for the superficial layer to capture permanent displacements. Boundary conditions were set for the model to ensure accurate results, and eight-node liner brick elements were used to mesh all pavement layers.

3 Results and Discussion In order to investigate the performance of pavement system, a series of 3-D-finite element dynamic simulations were carried out in order to evaluate the benefits offered against permanent deformation or rutting in function of the number of cycles of load.

36

D. J. Joseph et al.

3.1 Analysis of Pavement Section Three combinations of pavement types are analysed, firstly standard bitumen versus silica-gel modified bitumen and aerogel-incorporated bituminous pavement followed by standard concrete surface pavement with aerogel-incorporated concrete pavement, both combinations show the same trend. Both strain and deformation increased in modified pavements, while stress and reaction force decrease. These, both, results are used to compare the behaviour of aerogel-incorporated bitumen pavement. Different models of aerogel-incorporated bitumen pavement were created varying the aerogel percentages from 2.5 to 20%. Unlike the above models, the aerogelincorporated bituminous pavement shows an increase in all the four results, namely stress, reaction force, strain and deformation. Increase in the values of strain and deformation was as expected showing that virgin aggregates show the better results, while the increase in values of stress and reaction force shows that aerogel can enhance the grain-to-grain transfer of load due to the proper filling of voids than standard pavement. From all the different percentages of aerogel used, 2.5% and 5% give the best result as the deformation was comparable to the standard bitumen pavement and even less than the 5% silica-gel modified bitumen pavement.

3.2 Aerogel-Incorporated Bituminous Pavement Strain is increased compared to standard bitumen pavement by 13.44% and 21.59%, respectively, while strain in silica-gel pavement is increased by 62.26% compared to standard bitumen pavement. Thus, it can be inferred that the aerogel-incorporated pavements perform well compared to silica gel-incorporated pavement (Figs. 2 and 3). Stress is increased compared to standard bitumen pavement by 9.1% and 14.86%, respectively. Stress in silica-gel pavement is increased by 24.42% compared to standard bitumen pavement. Thus, aerogel-incorporated pavements perform better compared to silica gel-incorporated pavement (Table 4).

4 Conclusions This study is mainly focused on the use of silica aerogel in bituminous mix for a better road surface layer. The use of aerogel in road surfacing results in a more longer life span as the aerogel removes water from the pavement structure and provides thermal insulation, which results in improving the pavement properties. From the FEM analysis, aerogel-incorporated bitumen pavement is found to give better performance compared to silica-gel-modified bitumen pavement. Even though the strength of AIBP is slightly reduced, it performs well like the standard flexible

Comparative Studies on Gel-Incorporated Flexible Pavement

37

Fig. 2 Strain distribution between 2.5% (top) and 5% (bottom) aerogel pavement

pavement along with reduction of adverse effects of water and temperature changes due to the properties of aerogel. Further research must be performed to fully evaluate all the performance characteristics of aerogel-incorporated pavement.

38

D. J. Joseph et al.

Fig. 3 Stress distribution between 2.5% (top) and 5% (bottom) aerogel pavement

Comparative Studies on Gel-Incorporated Flexible Pavement

39

Table 4 FEM analysis results (variation in characteristics in comparison with conventional pavements) Strain variation (%)

Reaction force variation (%)

SGBP

62.26

14.86

24.42

26.53

AICP

48.83

8.96

10.03

24.86

Materials

Stress variation (%)

Deformation variation (%)

AIBP 2.5%

13.44

30.63

9.11

8.94

AIBP 5%

21.59

36.97

14.86

16.86

AIBP 7.5%

27.73

41.74

19.23

24.06

AIBP 10%

32.81

45.69

22.86

30.86

Acknowledgements We wish to express our sincere thanks to the APJ Abdul Kalam Technological University for providing all the support throughout the completion of the study.

References 1. Dhurai V, Padmakumar R (2021) Effect of joint width and sealing material on performance of bituminous block pavement using finite element method. Int J Eng Res Technol 9(6):37–41 2. Ganobjak M, Brunner S, Wernery J (2019) Aerogel materials for heritage buildings: materials, properties and case studies. J Cult Herit 3. Gao T, Jelle BP, Gustavsen A, Jacobsen S (2014) Aerogel-incorporated concrete: an experimental study. Constr Build Mater 52:130–136 4. Montes S, Maleki H (2020) Aerogels and their applications. In: colloidal metal oxide nanoparticles, pp 337–399 5. Lorenz Ratke (2011) Aerogels: structure, properties and applications. In: Conference: 13th meeting on supercritical fluids, ISASF 2011 6. Sanchez-Alonso E, Vega-Zamanillo A, Calzada-Perez MA, Castro-Fresno D (2018) Mechanical behaviour of asphalt mixtures containing silica gels as warm additives. Mate r Struct 51(4) 7. IRC:37-2018 Guidelines for the design of flexible pavements, Indian Road Congress, New Delhi. 8. IRC:58-2015 Guidelines for the design of plain jointed rigid pavements for highways, Indian Road Congress, New Delhi 9. IRC SP:53-2010 Guidelines on use of modified bitumen in road construction, Indian Road Congress, New Delhi 10. IRC SP:098-2013 Guidelines for the use of waste plastic in hot mix asphalt, Indian Road Congress, New Delhi

Performance Assessment of Premix Carpet for Low-Volume Roads Nishant Bhargava , Anjan Kumar Siddagangaiah , and Teiborlang Lyngdoh Ryntathiang

Abstract The paper presents the performance of premix carpet for low-volume roads in terms of structural capacity and functional properties. The structural capacity was evaluated using falling weight deflectometer (FWD). International roughness index (IRI) was used as a measure of the functional property. A total of 26 sections with a total road length of 110 km were selected for the study. At each test location, pavement composition was determined by excavating test pit along the edge of the shoulder. Subsequently, FWD test was conducted. Deflections were measured at 8 radial distances, ranging from 0 to 1500 mm. Deflection values ranged between 475 and 1129 µm below the loading plate and 30–224 µm at radial distance of 1500 mm. Using deflection readings, structural number (SN) and deflection bowl parameters including surface curvature index (SCI), base damage index (BDI), base curvature index (BCI) and AREA were calculated. Then, IRI was measured using Roughometer III device. IRI progression with time indicated that for a trigger value of 4.62 m/km, premix carpet roads would require maintenance after 41 months of service life. In addition, a good correlation between the deflection bowl parameters and IRI was observed. So, the limits of IRI were used to propose the recommended range of deflection bowl parameters. It was found that when the values of SCI, BDI and BCI increase to more than 313, 152 and 58, respectively, or AREA and SN reduce to less than 447 and 17, respectively, the pavement exhibits poor condition and would require rehabilitation. Keywords Premix carpet · Low-volume road · Falling weight deflectometer · IRI

N. Bhargava Birla Institute of Technology and Science Pilani, Pilani 333031, India A. K. Siddagangaiah (B) · T. L. Ryntathiang Indian Institute of Technology Guwahati, Guwahati 781039, India e-mail: [email protected] T. L. Ryntathiang e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_4

41

42

N. Bhargava et al.

1 Introduction Premix carpet (PMC) is widely used in India for the construction of low-volume roads. It comprises of aggregates having nominal size 13.2 and 11.2 mm, which are pre-mined with binder and compacted to a thickness of 20 mm. Study on low-volume roads constructed with PMC in different states of India showed that its performance after 2–5 years of service life was satisfactory. In this study, the distresses that were targeted include bitumen-rich/dry surface, aggregate content less/excessive, surface texture, cracking, potholes, surface evenness, raveling and stripping. Minimal distresses were observed during the test period [1]. However, limited studies had addressed the issue related to development of maintenance plan for PMC. In this regard, there is a need to assess the performance of low-volume roads. One of the most commonly adopted non-destructive test for structural evaluation of pavements is FWD. In this test, an impulse loading device in which a transient load simulating the actual traffic load is applied on the pavement surface. A fixed mass is dropped from a predetermined height on the system of springs placed over a circular loading plate. The corresponding peak load and peak deflection are measured at different radial distances. These deflections are then used to compute elastic moduli of pavement layers [2]. Different deflection bowl parameters including surface curvature index (SCI), base damage index (BDI), base curvature index (BCI) and AREA [3] have also been utilized to quantify the structural condition index of the pavement. Another key parameter of low-volume road performance is its riding quality. International Roughness Index (IRI) is generally used to assess the riding quality. IRI is the measure of undulations along the pavement surface, i.e., total rise and fall with respect to true planar surface. Higher IRI often leads to poor riding quality, higher vehicle operation cost and increased risk of road accidents. One of the commonly used technique is accelerometer-based measurement. In such technique, the vertical acceleration is measured using sensor fitted at the rear axle of the test vehicle while the distance is measured using distance measure instrument. However, studies on utilization of the deflection bowl parameters to determine the structural condition index rating of the pavement have been limited to roads constructed with surface layer of more than 40 mm thickness. In addition, limited studies have determined the IRI of low-volume roads. Hence, field investigations on structural capacity and functional properties could provide the much-needed information regarding the performance of low-volume roads.

1.1 Objective The objective of the study is to assess the performance of premix carpet for lowvolume roads in terms of structural capacity and functional properties. The following tasks were assigned for fulfilling the research objective.

Performance Assessment of Premix Carpet for Low-Volume Roads

43

. Identification of test sections and determination of pavement composition. . Evaluation of structural capacity using FWD test and calculation of deflection bowl parameters from the deflection data. . Assessment of pavement roughness in terms of IRI. . Determination of association between structural condition index parameters and IRI.

2 Methodology The research methodology adopted in the study was broadly divided into four phases: . Phase 1: Selection of test sections The study was conducted in the two states of Northeast India including Assam and Meghalaya constructed under Pradhan Mantri Gram Sadak Yojana (PMGSY) initiative. Overall, 26 projects were selected in this study (project P1 to P26), out of which 17 projects were from Assam and 9 were from Meghalaya. The total length of all projects combined was 110 km, where 56 km length was from Assam and the remaining 54 km length was from Meghalaya. The terrain for Assam region was plain/rolling for all sections. For Meghalaya, the terrain was mountainous for all sections. . Phase 2: Data collection For each section, at least one test pit having length and breadth of 0.6 m and depth upto subgrade were excavated, as shown in Fig. 1. The pavement composition was determined for each section. . Phase 3: Structural strength assessment Falling weight deflectometer (FWD) was used to evaluate the structural integrity of the pavement. The test equipment is shown in Fig. 1. Given that the pavement condition for the test sections was good to fair, the FWD measurements were

(a) Fig. 1 Photos illustrating a Test pit and b FWD equipment

(b)

44

N. Bhargava et al.

conducted at the interval of less than 130 m. Deflection data were then used to calculate the deflection bowl parameters and structural number. . Phase 4: Roughness measurement Pavement roughness was measured using Roughometer III. The results were computed in terms of International Roughness Index (IRI). Then, IRI was correlated with the structural number and deflection bowl parameters to benchmark the FWD deflection bowl parameter values for low-volume roads constructed with 20 mm thick PMC.

3 Experimental Protocols 3.1 Test Pit Excavation In this study, one test pit was excavated for every 2 km of road section. Each test pit was excavated along the outer lane of the earthen shoulder such that the pavement layers were exposed. The dimension of the test pit was 0.6 m × 0.6 m with depth upto subgrade layer. The interface between two pavement layers was identified visually, and the data regarding number of layer and layer thickness were recorded [2].

3.2 Falling Weight Deflectometer In this study, for the measurement of deflection using FWD, the plate load along with frame having displacement transducers was lowered to the test point on the pavement. The weight was raised to a predetermined height such that the load applied on the pavement was 40 kN. The load was then dropped, and the peak load and deformation were recorded. A total of eight displacement transducers placed at radial distances of 0, 200, 300, 450, 600, 900, 1200 and 1500 mm were used to measure deflection. At each test point, four readings were taken, out of which the first was the seating load and the next three readings were considered for further analysis [2].

3.3 International Roughness Index Roughness of the pavement is a key performance indicator, especially for low-volume roads. It is a measure of surface irregularities, i.e., sum of rise and fall with respect to the selected datum. Higher roughness has an adverse effect on the riding quality along with increased vehicle operation cost. Pavement roughness is generally defined in terms of IRI. In this study, the ARRB Roughometer III, which is designed to provide roughness data for both sealed and unsealed roads, was used to assess the

Performance Assessment of Premix Carpet for Low-Volume Roads

45

performance of the selected road sections. It uses a combination of wheel-mounted motion sensor and a distance input to measure the true longitudinal profile of the road. The longitudinal profile is used to calculate the IRI. During field investigations, the various components of the Roughometer III were initially attached to the survey vehicle as per the equipment standards. After installation of all the components, the test vehicle was taken to the start of the project. The vehicle was allowed to accelerate and reach a constant speed at the start of the road section. The survey was carried out within the speed limit of 30–60 kmph. As soon as the vehicle was able to maintain a constant speed between the limit, the survey was started. All the data collected were stored in the hand-held device, which was processed using a desktop software supplied along with the equipment.

4 Results and Discussions 4.1 Pavement Composition The pavement composition determined using the test pit excavation is shown in Fig. 2. It could be observed that the pavement composition for all section has premix carpet of 20 ± 2 mm thickness as wearing course. The base and sub-base layer comprised of water bound macadam (WBM) and granular sub-base (GSB), respectively. The thickness of WBM ranged from 75 to 150 mm whereas for GSB, the layer thickness varied from 100 to 200 mm.

4.2 Falling Weight Deflectometer FWD was conducted on the road sections having good to fair conditions only. It was observed that some of the road sections were damaged while some sections were inaccessible to carry the FWD equipment. So, FWD measurements were conducted on 17 sections only. The range of deflection values observed in these 17 sections is shown in Table 1. It could be observed that the average deflection below the loading plate, i.e., D0 , was in the range of 475–1129 µm. Only two sections exhibited D0 greater than 1000 µm whereas the remaining sections had D0 less than 760 µm. This shows that the structural capacity of 15 out 17 sections is satisfactory. Deflection bowl Parameters The deflection readings from FWD test data were used to determine the deflection bowl parameters. In this study, four parameters including surface curvature index (SCI), base damage index (BDI), base curvature index (BCI) and AREA were used [3]. The formulas and the associated application are shown in Table 2. For all the formulas, Di is the deflection at the radial distance i mm.

46

N. Bhargava et al.

Fig. 2 Pavement composition determined after test pit excavation Table 1 Deflection data from FWD test Statistic

Deflections (µm) D200

D300

D450

D600

Minimum

D0 474.7

330.3

229.2

160.3

113.8

D900 66.5

D1200 48.2

D1500 29.9

Maximum

1128.7

783.9

539.8

339.1

217.6

126.9

112.6

224.0

Table 2 Deflection bowl parameters [3] Serial number

Parameter and relation

1

Surface curvature index SCI = D0 − D300 Indication of primarily the base layer structural condition

2

Base damage index BDI = D300 − D600

3

Base curvature index BCI = D600 − D900 Indication of the lower structural layers like the selected and the subgrade layers

4

AREA method Normalized area under the deflection D300 basin from centre of load to radial +5× +6× AREA1200 = 4+6× DD200 D0 0 D450 D600 D900 D1200 distance of 1200 mm from the test load D0 + 9 × D0 + 18 × D0 + 12 × D0

Indication

Indication of the subbase and probably selected layer structural condition

Performance Assessment of Premix Carpet for Low-Volume Roads

800

500

600

400 BDI

SCI

300

400

200

200 0

100 0

Projects

800

150

600

Projects

AREA

BCI

200

100

400

50 0

47

200 0

Projects

Projects

Fig. 3 Deflection bowl parameters for different projects

The results of the deflection bowl parameters are shown in Fig. 3. The range of values of deflection bowl parameters observed is as follows: . . . .

SCI = 250–600 µm BDI = 115–320 µm BCI = 40–100 µm AREA = 400–550

Study on limits developed for Indian conditions suggested that the maximum permissible value for the warning condition for SCI, BDI and BCI is 600, 250 and 125, respectively [4]. The comparison of test results with the specification limits showed that almost all the test sections were having structural condition rating of good to warning. However, since the values were developed with pavement composition having 80–100 mm thick bituminous layer and 750 mm granular layer, it could be said that the limits should be relaxed even more for low-volume roads. Structural Number Structural Number (SN) is often used as an indicator for structural condition index. In this study, Eq. 1 was used to determine the SN for all test sections [5]. SN = a0 SIPa1 HPa2

(1)

where SIP = Structural Index of Pavement = D0 − D1.5HP = D0 − D450 ; HP = Total pavement thickness, mm; coefficients a0 , a1 , a2 were assigned as 0.1165, −0.3248

48

N. Bhargava et al.

IRI, m/km

Fig. 4 IRI progression with time

7 6 5 4 3 2 1 0

y = 3.07e0.01x

0

20

40 Age, months

60

80

and 0.8241, respectively [5]. For the 17 test sections, the SN varied within the range of 13.8–18.7, with the average value being 16.2.

4.3 International Roughness Index In this study, the pavement roughness was evaluated using Roughometer III. The results were quantified in terms of IRI. Then, utilizing the data collected, the riding quality deterioration with time was determined. To develop the relation, the traffic conditions were considered similar on all project roads. The road sections were grouped into five categories 12 months old, 12–24 months, 24–36 months, 36– 48 months and 48–60 months (or higher). Few outliers were removed to analyse the data considering traffic and soil variations and selection bias. Results illustrated in Fig. 4 showed that there was an exponential increase in IRI with time. For a trigger level of 4.62 m/km of IRI (3500 mm/km Unevenness Index) [6], the model pointed out that low-volume roads constructed using premix carpet would require an improvement in riding quality at the end of 41 months (almost 3.5 years).

4.4 Relation Between Structural Condition Index and IRI Parameters for structural condition index including SCI, BDI, BCI, AREA and SN were correlated with IRI. Results of the analysis are shown in Fig. 5. It could be observed that IRI increases exponentially with the increase in SCI, BDI and BCI. On the other hand, with the increase in AREA and SN, IRI decreases exponentially. This could be explained by the fact that lower values of SCI, BDI and BCI represent better structural condition of base, sub-base and subgrade layer. With the increase in the values of these parameters, the structural condition deteriorates leading to a

Performance Assessment of Premix Carpet for Low-Volume Roads

8

R² = 0.53

6

IRI, m/km

IRI, m/km

8 4 2 0 8

200

400 SCI

600

4 2 100

200 BDI

8

R² = 0.56

6

0

800

IRI, m/km

IRI, m/km

R² = 0.45

6

0

0

4 2

300

400

R² = 0.56

6 4 2 0

0 0

50

BCI

100

150

8 IRI, m/km

49

6 4 R² = 0.70

2 0 12

15

SN

18

21

350

400

450 500 AREA

550

Trigger level IRI = 4.62 m/km SCI = 313 BDI = 152 BCI = 58 AREA = 447 SN = 17

Fig. 5 Relation between IRI and structural condition index parameters

decline in the riding quality. Alternatively, higher SN represents a relatively better structural condition of the pavement. Hence, lower IRI was observed. After establishing the relationship, the threshold limits of structural condition index parameters were determined. The trigger level of IRI was considered as 4.62 m/ km, according to IRC recommendations [6]. Corresponding to IRI of 4.62 m/km, the threshold values of SCI, BDI, BCI, AREA and SN using the equation of the model. Result of the analysis is shown in Fig. 5. The proposed threshold limits could be utilized for effectively planning the maintenance and rehabilitation activities for low-volume roads. It should be noted that these threshold limits were developed with the limited dataset. Different climatic conditions, traffic composition and material properties should be considered for generalizing the proposed model.

50

N. Bhargava et al.

5 Conclusion This study evaluated the performance of 20 mm thick premix carpet for low-volume roads. A total of 26 sections were selected, out of which 17 sections were tested for structural capacity and functional property. Structural condition index using deflection measurements from FWD test and roughness measurement using IRI were used to characterize the performance. The following conclusion could be drawn from this study. . Structural condition index rating was within the range of sound to warning for most of the test sections. It shows that premix carpet performs well for low-volume roads. . IRI increased exponentially with time. The model predictions pointed out that the low-volume roads would require a riding quality improvement after 3.5 years of service life, considering the trigger level of IRI as 4.62 m/km. . The deflection bowl parameters and structural number showed a good correlation with IRI. Hence, the quality of subgrade, sub-base and base construction can influence the progression of riding quality with time in low-volume roads. . Threshold limits of SCI, BDI, BCI, AREA and SN were proposed in the study, considering the trigger level of IRI as 4.62 m/km. The outcomes of the study were established based on the limited dataset available. Further investigations could help in generalizing the established model and the threshold limits proposed in this study. Acknowledgements Authors would like to acknowledge National Rural Infrastructure Development Agency (NRIDA), Ministry of Rural Development, Government of India for providing financial support. Authors also appreciate Public Works (Roads), Government of Assam and Meghalaya for their help in conducting the field investigations.

References 1. Pundhir NKS, Nunda PK (2006) Development of bitumen emulsion based cold mix technology for construction of roads under different climatic conditions of India. J Sci Ind Res 65:729–743 2. IRC 115: Guidelines for structural evaluation and strengthening of flexible road pavements using falling weight deflectometer (FWD) technique. Indian Roads Congress, New Delhi, India (2014) 3. Pierce LM, Bruinsma JE, Smith KD, Wade MJ, Chatti K, Vandenbossche JM (2017) Using falling weight deflectometer data with mechanistic-empirical design and analysis, volume III: guidelines for deflection testing, analysis, and interpretation, Report No. FHWA-HRT-16-011. Federal Highway Administration, Washington, DC 4. Mosale Ramanath A, Venkateshappa A, Macheri Srinivasarao A, Amirthalingam V (2020) Benchmarking falling weight deflectometer deflection bowl parameters: case study for Indian conditions and the applications to rehabilitation design. J Transp Eng Part B: Pavements 146(3):05020003

Performance Assessment of Premix Carpet for Low-Volume Roads

51

5. Rohde GT (1994) Determining pavement structural number from FWD testing. Transp Res Rec 1448:61–68 6. IRC SP 16: Guidelines on measuring road roughness and norms (second revision). Indian Roads Congress, New Delhi, India (2019)

Development of Resilient Modulus Model for the Bituminous Course Paras Markana, Bharath Gottumukkala, Akshay Gundla, Ambika Behl, and Tejaskumar Thaker

Abstract As per IRC 37, the design of flexible pavement requires the accurate prediction of resilient modulus for bituminous concrete. This, determination of Mr requires specialized test equipment, which may not be available in many laboratories. Therefore, it would be rather reasonable to create a model that could estimate Mr from easy to estimate volumetric parameters and other parameters obtained from sample conventional tests. This study aims to investigate the effect of different binders (VG30, VG40, PMB and CRMB), binder contents, volumetric parameters (air voids) and temperature on Resilient Modulus (Mr ) in Bituminous Concrete (BC) course and Dense Bituminous Macadam (DBM) course. The samples were compacted using the Marshall Compacter and subsequently the volumetric parameters were calculated. The Indirect Tensile Strength (ITS) test was conducted to calculate the ITS and Toughness. The Repeated Load Indirect Tension Test was conducted to find out the Resilient modulus (Mr ) values for all mixes. The regression analysis was performed using the Microsoft Excel tool, and the relationship between resilient modulus, volumetric parameters and ITS was developed. Using the final model, Resilient Modulus of mixtures may be predicted from volumetric parameters and indirect tensile strength under comparable or different testing conditions. Keywords Bituminous concrete (BC) course · Dense bituminous macadam (DBM) · Resilient modulus · Indirect tensile strength · Polymer modified bituminous (PMB) mixes · Crumb rubber modified bituminous (CRMB) mixes P. Markana (B) Transportation Engineering, Pandit Deendayal Energy University, Gandhinagar 382007, India e-mail: [email protected] B. Gottumukkala · A. Gundla · A. Behl CSIR-Central Road Research Institute, New Delhi 110025, India e-mail: [email protected] A. Behl e-mail: [email protected] T. Thaker Department of Civil Engineering, Pandit Deendayal Energy University, Gandhinagar 382007, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_5

53

54

P. Markana et al.

1 Introduction The quality of a country’s road transportation system has a significant effect on its economic and social growth. The majority of India’s road network is made up of flexible pavements [1]. The design of flexible pavement requires accurate prediction of resilient modulus of all the layers of pavement. If resilient modulus predictions are inaccurate, pavement will fail before it is intended to, costing a lot of money to repair [2]. Under repeated loading tests, the resilient modulus is defined as the ratio of deviator stress to recoverable elastic strain [3]. The temperature environment to which flexible pavements are subjected has a significant impact on their functional and structural performance. Temperature has been demonstrated to influence the strength and deformation properties of asphalt concrete mixtures, affecting the resilience modulus of the asphalt layers [4]. The asphalt type (unmodified and modified asphalt) and aggregate gradation also affect the resilient modulus. Resilient modulus changed as a result of the asphalt binder and aggregate gradation changes [5]. Considering the issues connected with cracking, the tensile characteristics of asphalt mix are of excessive interest to pavement engineers. As a result, the tensile strength of asphalt concrete mix is critical for asphalt paving. The indirect tensile strength test is performed to evaluate the bituminous mix’s tensile qualities [6, 7]. The comparative study of ITS test was conducted on modified asphalt mixes [8]. The harshness of the highway system’s problems has been compounded by increased traffic loads, necessitating the improvement in the quality of existing bituminous materials. Polymer binder modification has been offered as a solution to improve the physical and rheological behaviours of bituminous binder [9]. The polymer modified asphalt concrete has a greater resilience modulus than the unmodified asphalt concrete. Polymer modified asphalt concrete gives higher strength and durability than the unmodified asphalt concrete [10]. Crumb Rubber Modified Bitumen (CRMB) is also one of the option other than the polymer modified bituminous mixes. Crumb rubber is a recycled rubber made mostly from natural, synthetic, and carbon black rubbers found in automobile trash tyres [11]. Crumb rubber enhances the viscosity, softness point, and temperature susceptibility of modified bitumen, as well as its resistance to deformation at increasing pavement temperatures [12]. Bituminous mix design is a difficult procedure that needs precise proportions of aggregate and asphalt binder in order to meet particular volumetric and mechanical criteria [13]. A good performance-volumetric relationship (PVR) can help mixes work more effectively and may be employed in future quality design mixes. Voids in mineral aggregate (VMA) are a critical design parameter in the current Superpave mix design technique that relates hot-mix-asphalt (HMA) mix qualities to field performance. The ability to achieve the proper VMA early in the mix-design process has a considerable impact on design time and effort, as well as the ability to generate a cost-effective and appropriate design with acceptable field performance. Air voids is also one of the volumetric parameters, which affects the performance of the pavement [14]. To prevent binder ageing, permeability, and resultant stripping difficulties, asphalt mixes should contain the least possible air spaces [15]. The addition

Development of Resilient Modulus Model for the Bituminous Course

55

Table 1 Resilient modulus predictive models Model name ANN model

Resilient modulus for wearing course

Equation    Ep = 21 j dpj − Opj  E = p Ep

[17]

Mr = −1297074.90 − 165520594 T − 52354.064 PS

[18]

−15094.311 AV + 301011.367 η + 2306479.544 SA

Model for using waste PET plastic modified bituminous mixes

Mr = 1.08102 ∗ (ITS)1.124 − 465 ∗ T

[19]

Model for WMA

Mr (psi) = 915411 + 19343 (AC) − 7928.4 (TT)

[20]

−282.24 (LD) + 39540 (Filler Type) − 89014 (Av )

of polymer and crumb rubber in unmodified bitumen gives more accurate results for the percentage of voids in mineral aggregate and percentage voids field with the asphalt [16]. Various regression models that are already developed to find out resilient modulus value are listed in Table 1.

2 Objectives of the Study The primary objective of the study was to a resilient modulus model that relates Mr to the volumetric parameters of bituminous mixtures.

3 Experimental Design and Procedure 3.1 Materials In this study, the experimental design included the utilization of aggregate, four different types of binders (VG30, VG40, PMB and CRMB) and stone dust as filler material. For the preparation of Marshall samples, BC-1 and DBM-1 gradation were used as per MoRTH specification. The sample was prepared using the mid-point gradation. The aggregate’s engineering properties were measured and compared to the respective specifications given in Table 2. The engineering properties of the binders are measured and compared to the corresponding standards. The testing was performed on all four binders as per the procedures described in the relevant codes, and the findings are shown in Tables 3 and 4.

56

P. Markana et al.

Table 2 Engineering properties of the aggregate Characteristics

Obtained

Limit

Specification

Aggregate impact value (AIV)

11.83%

Max 24%

IS:2386 (Part IV)

Combined flakiness and elongation indices

30.29%

Max 35%

IS:2386 (Part I)

Specific gravity

2.77

–

ASTHO T84 & T85

Water absorption

0.2%

Max 2%

IS:2386 (Part III)

Table 3 Engineering properties of VG30 and VG40 grade bitumen Characteristics

VG30 Obtained

VG40 Limit

Obtained

Specification Limit

Specific gravity

1.01

–

1.03

–

IS:1202

Penetration at 25 °C, 100 g, 5 s, 0.1 mm, min

52

45

49

35

IS:1203

Softening point, °C, min

51

47

53.2

50

IS:1205

Kinematic viscosity, 135 °C, poise

445

350

525

400

IS:1206 (Part 1)

Failure temperature °C

79

–

88

–

AASHTO T315

G*/sinδ, kPa

1.00

≥1

1.15

≥1

AASHTO T315

Table 4 Engineering properties of PMB and CRMB binders Characteristics

PMB Obtained

CRMB Limit

Obtained

Specification Limit

Specific gravity

1.03

–

1.01

–

IS:1202

Penetration at 25 °C, 100 g, 5 s, 0.1 mm, min

33.3

30–50

58.5

50

IS:1203

Softening point, °C, min

81.3

60

67

60

IS:1205

Viscosity, 150 °C, poise

4

3–9

5.25

3–9

IS:1206 (Part 1)

Failure temperature °C

94

–

94

–

AASHTO T315

G*/sinδ, kPa

1.32

≥1

1.51

≥1

AASHTO T315

3.2 Experimental Program This experimental study includes four types of binder (VG30, VG40, PMB and CRMB). The aggregates and stone dust were locally sourced. To verify that the materials conform to the minimal standards specified in the codal regulation, physical properties of all binders and aggregates were verified. HMA specimens were prepared using four different binders and were compacted using a Marshall compactor. BC-1 (at 4.5 and 5.5% binder contents) and DBM-1 (at 4 and 6% binder contents) gradations were utilized to prepare the test samples. Please note that the binder contents used

Development of Resilient Modulus Model for the Bituminous Course

57

in this study are under and above the anticipated OBC, this was done in order to look at bitumen content sensitivity to Mr so that wider differences in properties can be achieved. Following the preparation of the specimen, density-void analysis was performed to determine all volumetric characteristics. Following the volumetric investigations, an indirect tensile strength (ITS) test and a resilient modulus (Mr ) test were performed at 25 and 35 °C. The main objective of this research is to develop a model to predict the resilient modulus using ITS results and the mixture volumetric properties.

4 Results and Discussions 4.1 Viscosity Test Rotational viscosity test was performed using a Brookfield viscometer in accordance with ASTM D 4402. This test technique is used to assess the apparent viscosity of asphalt binders at handling, mixing or application temperatures. The results of the viscosity test are shown in Table 5.

4.2 Mixing and Compaction Temperatures According to MS-2, equiviscous temperature ranges were employed in asphalt mix design procedures for laboratory mixing and compaction temperatures. Equiviscous mixing and compaction temperatures are utilized in laboratory mix design approaches to normalize the influence of asphalt binder stiffness on mixture volumetric characteristics. Using that viscosity value, the mixing and compaction temperatures for all four binder mixes were calculated using the equiviscous principle. Table 6 shows the mixing and compaction temperatures of all binder mixtures. Table 5 Viscosity of binders

Binder

Viscosity (Pa.S) 135 °C

150 °C

170 °C

VG30

0.450

0.100

–

VG40

0.525

0.225

0.075

PMB

0.825

0.400

0.150

CRMB

1.050

0.525

0.225

58

P. Markana et al.

Table 6 Mixing and compaction temperature of mix Binder

Mixing temperature (°C)

Compaction temperature (°C)

VG30

152–155

142–147

VG40

158–161

148–153

PMB

166–168

159–163

169.5–171

164–167

CRMB

4.3 Volumetric Study Asphalt mixtures’ volumetric characteristics are utilized to attain a satisfactory performance. The volumetric parameters were calculated using the MS-2 equations and the results are shown in Table 7. Table 7 Volumetric properties of Mix Gradation and binder content

Volumetric parameters

BC-1 (5.5% binder content)

(Pa )

BC-1 (4.5% binder content)

VG30

VG40

PMB

2.38

2.77

2.10

2.10

VMA (%)

11.98

9.98

9.77

9.64

VFA (%)

78.23

80.16

72.27

78.51

Gse

2.91

2.92

2.90

2.91

Pba (%)

2.76

2.84

2.64

2.73

Pbe (%)

2.89

2.82

3.01

2.92

(Pa )

3.07

3.33

2.56

2.08

VMA (%)

11.29

11.14

10.54

10.82

VFA (%)

72.82

70.17

75.78

80.92

2.87

2.88

2.88

2.86

Gse

DBM-1 (4% binder content)

Pba (%)

1.30

1.44

1.40

1.14

Pbe (%)

3.26

3.12

3.17

3.41

(Pa )

3.71

3.64

3.50

3.20

VMA (%)

11.63

12.33

11.20

11.71

VFA (%)

68.21

70.69

73.36

72.70

2.84

2.81

2.83

2.85

Gse

DBM-1 (6% binder content)

CRMB

Pba (%)

1.82

1.44

1.70

2.02

Pbe (%)

2.25

2.62

2.36

2.07

(Pa )

1.44

1.88

1.27

1.09

VMA (%)

11.10

11.47

11.29

10.67

VFA (%)

87.29

83.61

83.92

90.05

2.78

2.78

1.17

2.80

Gse Pba (%)

1.12

1.10

1.17

1.35

Pbe (%)

2.93

2.95

2.88

2.70

Max. Strength (MPa)

Development of Resilient Modulus Model for the Bituminous Course

59

2.500 2.000 1.500 1.000

4.5 % bc, BC-1

0.500 0.000

5.5 % bc, BC-1

4.5 % bc, BC-1

VG30 0.827

VG40 0.688

PMB 0.950

CRMB 0.861

5.5 % bc, BC-1

0.921

0.849

1.099

0.983

4 % bc, DBM-1

1.215

1.020

1.931

1.066

6 % bc, DBM-1

0.681

0.698

1.055

0.838

4 % bc, DBM-1 6 % bc, DBM-1

Binders

Fig. 1 Indirect tensile strength of mix

4.4 Indirect Tensile Strength (ITS) In conjunction with laboratory mix design testing, the Indirect Tensile Strength (ITS) test may be performed to assess the relative quality of bituminous mixes and predict the possibility of rutting or cracking. The ASTM D 6931-12 guideline was followed for performing the indirect tensile strength (ITS) test. The results of the indirect tensile strength test are shown below. The results demonstrate that PMB mixes have the highest indirect tensile strength when compared to the other binder mixes at both the binder content and both the gradation BC-1 and DBM-1, as shown in Fig. 1. The peak load and corresponding deformation values were provided by ITS test. The stress and strain were determined using the load and deformation data. Toughness was then estimated from the stress–strain curves. Toughness is generally defined as the area under the stress–strain curve, and it also represents the energy absorption capacity. Figure 2 illustrates the toughness values of all binder mixtures using the BC-1 and DBM-1 gradations. The results reveal that PMB mixes have the highest toughness value when compared to other binder mixes at both binder contents and considering both BC-1 and DBM-1 gradations.

4.5 Resilient Modulus (Mr ) Resilient modulus is used in the evaluation of material quality and as input for pavement design, evaluation and analysis. Repeated load indirect tension test is used to find the resilient modulus value. Resilient modulus Mr test has been conducted as per the ASTM D 4123 guideline. At 25 and 35 °C, 100 cycles of load repetition

P. Markana et al.

Area under the StressStrain curve (N/m2)

60

0.60 0.50 0.40 0.30 0.20 0.10 0.00

4.5% bc, BC-1

VG30 0.17

VG40 0.14

PMB 0.32

CRMB 0.21

5.5% bc, BC-1

0.23

0.16

0.47

0.31

4% bc, DBM-1

0.23

0.15

0.29

0.24

6% bc, DBM-1

0.19

0.21

0.53

0.38

4.5% bc, BC-1 5.5% bc, BC-1 4% bc, DBM-1 6% bc, DBM-1

Binders

Fig. 2 Area under the stress–strain curve

Mr (MPa)

are applied. In the resilient modulus test, 10% of the ITS pick load was used as cyclic load and 4% of cyclic load was used as contact load. Figures 3 and 4 show the resilient modulus results of the various mixtures. The results of the resilient modulus test demonstrate that the PMB mixes had the highest resilient modulus at both temperatures. As the binder content falls, so does the resilient modulus value for BC-1 mixtures, however, opposite trend is seen for DBM mixtures. Both the BC-1 and DBM-1 gradations showed a similar tendency. Temperature is another key and significant component for the resilient modulus test. When the temperature increases, the value of the resilient modulus falls.

3500 3000 2500 2000 1500 1000 500 0

25 Ϲ, 5.5 % bc VG30 2549

VG40 1652

PMB 2651

CRMB 1358

35 Ϲ, 5.5 % bc

25 Ϲ, 5.5 % bc 35 Ϲ, 5.5 % bc

1213

1477

1867

815

35 Ϲ, 4.5 % bc

25 Ϲ, 4.5 % bc

2137

1305

2649

1727

35 Ϲ, 4.5 % bc

1352

1194

1432

1167

Binders

Fig. 3 Mr for BC-1 gradation

25 Ϲ, 4.5 % bc

Mr (MPa)

Development of Resilient Modulus Model for the Bituminous Course

7000 6000 5000 4000 3000 2000 1000 0

61

25 Ϲ, 4 % bc

25 Ϲ, 4 % bc

VG30 4084

VG40 1825

PMB 5480

CRMB 2059

35 Ϲ, 4 % bc

2470

1019

2095

1488

25 Ϲ, 6 % bc

1241

942

2085

1874

35 Ϲ, 6 % bc

690

695

1006

833

35 Ϲ, 4 % bc 25 Ϲ, 6 % bc 35 Ϲ, 6 % bc

Binders

Fig. 4 Mr for DBM-1 gradation

4.6 Regression Analysis The data for the regression analysis was acquired using volumetric investigations, Indirect Tensile Strength test, and Resilient Modulus test. Some of the parameters considered in the resilient modulus model are Toughness, Indirect Tensile Strength, Binder Content, Test Temperature, and Percentage Air Voids. Using the acquired data set, the Resilient Modulus Model for Bituminous Mixtures was developed by regression analysis using the Microsoft Excel tool, as shown below. Mr = 2980.99 − 395.91 (S) + 2124.14 (ITS) − 123.47 (bc) − 76.43 (T) − 155.985 (Av ) where Mr = Resilient Modulus (MPa) S = Toughness value (MPa) ITS = Indirect Tensile Strength (MPa) bc = Binder Content (%) T = Temperature (°C) Av = Air Voids (%) The resilient modulus model indicates that as the indirect tensile strength increases, so does the resilient modulus, however when the other factors such as toughness, binder content, temperature, and air voids increase, the resilient modulus decreases. The measured versus predicted resilient modulus, as well as the prediction accuracy of the analysed models for testing data sets, are shown in Fig. 5. The

62

P. Markana et al.

3500

Predicted Mr

3000 R² = 0.6309

2500 2000 1500 1000 500 0 0

500

1000

1500

2000

2500

3000

3500

Measured Mr

Fig. 5 Calculated versus predicted Mr

majority of the points are close to the line of equality indicates that the model produces appropriate results and may be used to calculate the resilient modulus value.

5 Conclusions The resilient modulus is an important consideration when designing flexible pavements. The primary objective of this study was to develop a predictive resilient modulus model for the bituminous mixture. The laboratory experiments comprised of the ITS test and the Mr test, which were performed at different test temperatures (25 and 35 °C) considering the BC-1 (4.5 and 5.5% bc) and DBM-1 (4 and 6% bc) gradations. The following conclusions are derived based on the test data obtained in the laboratory and analysis performed. • PMB mix has the highest ITS value. • PMB mix has the highest toughness in both gradations and at both binder contents. In both the BC-1 and DBM-1 gradations, similar trend is observed in toughness as the binder content was reduced. • The PMB mix of BC-1 gradation had the highest Resilient Modulus (Mr ) values at 25 and 35 °C. Whereas PMB mix of DBM-1 gradation had the highest Resilient Modulus values at both temperatures 25 and 35 °C when binder concentration was 6%, when binder content was reduced to 4%, PMB mix had the highest Resilient Modulus values at 25 °C and VG30 mix had the highest Resilient Modulus values at 35 °C. The value of the resilient modulus falls as the temperature increases. • The regression analysis revealed that temperature and indirect tensile strength had a significant influence on resilient modulus value.

Development of Resilient Modulus Model for the Bituminous Course

63

The developed resilient modulus model can be used to predict resilient modulus values from ITS testing and volumetric data, which will be useful for laboratories without specialized equipment to predict reasonably the values of resilient modulus.

References 1. Gautam PK, Kalla P, Jethoo AS, Agrawal R, Singh H (2018) Sustainable use of waste in flexible pavement: a review. Constr Build Mater 180:239–253. https://doi.org/10.1016/j.conbuildmat. 2018.04.067 2. Gottam SR, Adepu R, Penki R (2020) Evaluation of bituminous mix characteristics prepared with laboratory developed high modulus asphalt binder. J Inst Eng Ser A 101(4):701–712. https://doi.org/10.1007/s40030-020-00462-4 3. Behbahani H, Ayazi MJ, Moniri A (2017) Laboratory investigation of rutting performance of warm mix asphalt containing high content of reclaimed asphalt pavement. Pet Sci Technol 35(15):1556–1561. https://doi.org/10.1080/10916466.2017.1316738 4. Al Mamun A, Al-Abdul Wahhab HI, Dalhat MA (2020) Comparative evaluation of waste cooking oil and waste engine oil rejuvenated asphalt concrete mixtures. Arab J Sci Eng 45(10):7987–7997. https://doi.org/10.1007/s13369-020-04523-5 5. Radhakrishnan V, Dudipala RR, Maity A, Sudhakar Reddy K (2019) Evaluation of rutting potential of asphalts using resilient modulus test parameters. Road Mater Pavement Des 20(1):20–35. https://doi.org/10.1080/14680629.2017.1374994 6. Gupta L, Suresh G (2018) Determination of indirect tensile strength of bituminous concrete mix prepared using stone dust and cement as filler materials. Sustain Civ Infrastruct 1:249–261. https://doi.org/10.1007/978-3-319-61633-9_16 7. Purohit S, Panda M, Chattaraj U (2021) Use of reclaimed asphalt pavement and recycled concrete aggregate for bituminous paving mixes: a simple approach. J Mater Civ Eng 33(1):04020395. https://doi.org/10.1061/(asce)mt.1943-5533.0003480 8. Navarro FM, Gámez MCR (2012) Influence of crumb rubber on the indirect tensile strength and stiffness modulus of hot bituminous mixes. J Mater Civ Eng 24(6):715–724. https://doi. org/10.1061/(asce)mt.1943-5533.0000436 9. Yang X, You Z, Dai Q, Mills-Beale J (2014) Mechanical performance of asphalt mixtures modified by bio-oils derived from waste wood resources. Constr Build Mater 51:424–431. https://doi.org/10.1016/j.conbuildmat.2013.11.017 10. Enieb M, Shbeeb L, Asi I, Yang X, Diab A (2020) Effect of asphalt grade and polymer type (SBS and EE-2) on produced PMB and asphalt concrete mix properties. J Mater Civ Eng 32(12):04020385. https://doi.org/10.1061/(asce)mt.1943-5533.0003479 11. Hanumantharao C, Anil Pradhyumna T, Durga Prasad K, Naveenkumar N, Shantha Kumar Reddy G, Hemanth Vardhan M (2019) Crumb rubber modified bitumen and quarry dust in flexible pavements. Int J Recent Technol Eng 8(1):2868–2873 12. Bansal S, Kumar Misra A, Bajpai P (2017) Evaluation of modified bituminous concrete mix developed using rubber and plastic waste materials. Int J Sustain Built Environ 6(2):442–448. https://doi.org/10.1016/j.ijsbe.2017.07.009 13. Al-Bayati HKA, Tighe SL, Achebe J (2018) Influence of recycled concrete aggregate on volumetric properties of hot mix asphalt. Resour Conserv Recycl 130(November 2017):200–214. https://doi.org/10.1016/j.resconrec.2017.11.027 14. Wang YD, Ghanbari A, Underwood BS, Kim YR (2019) Development of a performancevolumetric relationship for asphalt mixtures. Transp Res Rec 2673(6):416–430. https://doi. org/10.1177/0361198119845364 15. Tahmoorian F, Samali B, Yeaman J, Mirzababaei M (2020) Evaluation of volumetric performance of asphalt mixtures containing recycled construction aggregate (RCA). Int J Pavement Eng 0(0):1–15. https://doi.org/10.1080/10298436.2020.1849686

64

P. Markana et al.

16. Gogoi R, Biligiri KP, Das NC (2016) Performance prediction analyses of styrene-butadiene rubber and crumb rubber materials in asphalt road applications. Mater Struct Constr 49(9):3479–3493. https://doi.org/10.1617/s11527-015-0733-0 17. Hu C (2011) Prediction of resilient modulus for hot mix asphalt based on artificial neural network. Adv Mater Res 304:18–23. https://doi.org/10.4028/www.scientific.net/AMR.304.18 18. Hilal M (2018) Prediction of resilient modulus model for wearing asphalt pavement layer. Kufa J Eng 09(4):65–87. https://doi.org/10.30572/2018/kje/090405 19. Modarres A, Hamedi H (2014) Developing laboratory fatigue and resilient modulus models for modified asphalt mixes with waste plastic bottles (PET). Constr Build Mater 68:259–267. https://doi.org/10.1016/j.conbuildmat.2014.06.054 20. Mahdi H, Albayati AH (2020) Model development for the prediction of the resilient modulus of warm mix asphalt. Civ Eng J 6(4):702–713. https://doi.org/10.28991/cej-2020-03091502

Forensic Investigations for Failure of Flexible Pavements: A Case Study Abhishek Mittal

and Amit Kumar

Abstract Flexible pavements undergo some form of distress during the design life. So, it becomes necessary to carry out forensic investigations to understand the causes of distresses/failures and develop an optimal rehabilitation strategy. The present study describes an investigation undertaken to determine the probable causes of distress in flexible pavements in an industrial township. The studies conducted include visual surface condition assessment, classified traffic volume counts, axle load studies, pavement deflection studies using Benkelman beam, bituminous core cutting, laboratory characterization of samples from all pavement layers, and full-depth pit cutting at some locations on the flexible pavements. During the condition assessment, it was observed that the condition of the pavement was poor on the entire stretch with heavy ravelling and potholes. From field and laboratory studies, it was inferred that all the pavement layers (including subgrade) were poorly compacted and did not meet the specified requirement as per the Indian specifications. Overall, the quality of the materials and workmanship was of poor quality. The pavement was heavily overloaded by commercial vehicles and a high vehicle damage factor (VDF) value was obtained. The deflection studies indicated that the existing pavement thickness was inadequate and need to be strengthened. Accordingly, suitable thickness of overlay was recommended to strengthen the pavement. Keeping in view the structural and surface condition of the existing pavement, traffic loads and damaging factors, the short-term and long-term corrective and rehabilitation measures were suggested. Keywords Failure investigations · VDF · Axle load · Benkelman beam

A. Mittal (B) · A. Kumar Flexible Pavement Division, CSIR-Central Road Research Institute (CSIR-CRRI), New Delhi 110025, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_6

65

66

A. Mittal and A. Kumar

1 Introduction Faridabad is a major industrial hub of Haryana. Haryana State Industrial and Infrastructure Development Corporation Ltd. (HSIIDC) is maintaining the roads in the industrial Model Township (IMT) area in Faridabad. The roads cater to heavy volumes of loaded trucks. The roads in the area were constructed during the period 2011 to 2016. The roads are in bad condition and need to be improved. Based on the request of HSIIDC, this study for the evaluation of the roads was taken up [1].

2 Objectives and Scope of the Study The main objective of this project is to carry out the investigation to study the causes of failure of the study road. The study included the evaluation of general road condition and structural condition in order to assess the deficiencies and to suggest suitable remedial measures. To achieve the above objectives, the scope of the study included the following activities: a) Field investigations • General condition assessment by visual survey • Structural evaluation of pavement by Benkelman Beam deflection measurements • Test pit observations and bituminous core sampling • Extraction of material samples from various pavement layers • Classified traffic volume studies and axle load studies b) Laboratory investigations by studying engineering properties of extracted pavement materials/mixes c) A comprehensive analysis, inferences and suggesting suitable remedial measures.

3 Field Studies The field investigation work was undertaken with a view to assess the quality of pavement layers and to carryout structural evaluation of the pavement, so that reasons of pavement failure could be made out.

Forensic Investigations for Failure of Flexible Pavements: A Case Study

67

Table 1 General condition of roads at IMT Faridabad Serial number

Road identification

1

Pavement surface condition w.r.t distress General road condition

Cracking

Ravelling

Rutting

Potholes

CETP plant to Poor Mohna road (90 wide road)

At isolated locations

At isolated locations

NIL

At isolated locations

2

Mohna distributary to entry at proposed bridge (90 m wide road)

Poor

At isolated locations

At isolated locations

NIL

At isolated locations

3

45 m wide road

Poor

At isolated locations

At isolated locations

NIL

At isolated locations

3.1 Visual Surface Condition Assessment The visual condition data were collected by observing different forms of distress on the road sections. It was observed that the roads have developed severe cracking and many potholes were observed. The summary of road condition assessment is given in Table 1.

3.2 Benkelman Beam Deflection Measurements To assess the structural condition of the road stretch under investigation, nondestructive method of Benkelman Beam rebound deflection has been used. Deflection measurement points covering the entire reach and representative of different conditions of study roads were selected for this study. The Benkelman Beam deflections were measured at points staggered at 50 m with a standard truck having rear axle load of 8.16 tonnes and tyre pressure of 5.6 kg/cm2 . The measurements were taken as per CGRA procedure laid down in IRC:81 [2].

3.3 Test Pit Observations Based upon the condition survey of the study area, the locations of the representative test pits were decided. A total of 11 test pits were identified on the various study roads. The test pits, measuring about 1.0 m by 1.0 m, were dug open upto the subgrade level at these identified locations to study the condition of constituent layers. Samples

68

A. Mittal and A. Kumar

from constituent layers were collected for further evaluation in the laboratory from all these test pits. The thickness of each constituent layer was measured in all the test pits and average value was taken. The measurements of in-situ densities of Wet Mix Macadam (WMM), Granular Sub-base (GSB) and subgrade soil were done by sand replacement and core cutter method.

3.4 Coring of Bituminous Layers (Surface and Binder Course) The bituminous layers of wearing course and binder course were examined by coring from the pavement to get samples of these mixes for evaluating their properties in the laboratory. Cores of bituminous mixes, of 100 mm diameter, were taken out using a core cutting machine at many different locations on the study roads. The core locations were staggered so as to take the cores from all the lanes of the roads. The cores were further separated out in the laboratory by slicing using automatic slicer for further characterization.

3.5 Estimation of Design Traffic For the purpose of design of strengthening/rehabilitation requirements, traffic, both in terms of both volume and axle loads, is an important parameter. Hence, present and the future projected traffic are considered for analysis in deriving the estimated load applications. Classified traffic counts and the ‘Vehicle Damage Factor’ are the requisites for traffic load estimation. Classified Traffic Volume Counts Classified traffic volume data were collected for 3 days (72 h) round the clock at the identified location on the study roads, for both the directions, covering all categories of vehicles. The traffic data were collected manually by trained enumerators under the continuous surveillance of CRRI team. The data on classified traffic counts for both directions of traffic are given in Table 2. The total number of commercial vehicles only, for each direction of travel, has been considered for overlay design. Axle Load Studies (Using Static Weigh Pads) The axle load studies were carried out near the main entry gate, for both the directions, for 24 h round-the-clock. The data obtained were processed to get the Vehicle Damage Factor (VDF), as given in subsequent section. VDF is a multiplier to convert the number of commercial vehicles of different axle loads and axle configurations to the number of standard axle load repetitions [3]. It is defined as the equivalent number

Forensic Investigations for Failure of Flexible Pavements: A Case Study

69

Table 2 Classified traffic volume data Direction of traffic

Towards CETP plant

Towards main entry gate

Vehicle type

Number

Percent composition of the total volume (%)

Number

Percent composition of the total volume (%)

Car/jeep

1844

41.9

2022

45.1

Light passenger vehicles/ mini-buses

4

0.1

7

0.2

14

0.3

25

0.6

Light commercial vehicles

724

16.5

877

19.5

Two axle trucks

791

18.0

813

18.1

Standard bus

Three axle trucks

426

9.7

298

6.6

Multi-axle trucks

384

8.7

275

6.1

Tractor/trailers

214

4.9

169

3.8

Total commercial vehicles 2539 for design (LCVs + trucks)

100

2432

100

Note The percentage (rounded-off values) indicated is in terms of number of vehicles in a particular category to the total number of all motorized vehicles

Table 3 VDF obtained from axle load study Direction of traffic Serial number

Vehicle type

1

Light commercial vehicles (LCV)

Towards CETP plant

Towards main entry gate

Individual VDF

Individual VDF

0.71

0.59

2

Two axle trucks

2.52

2.15

3

Three axle trucks

3.80

2.91

4

Multi-axle trucks

17.8

12.7

of standard axles per commercial vehicle. The VDF varies with the vehicle axle configuration, axle loading, terrain, type of road and from region to region (Table 3).

4 Laboratory Characterization of the Collected Materials The laboratory evaluation included testing for engineering properties of in-situ pavement materials. Laboratory evaluation of the materials collected from field aimed at evaluating them for their compliance with the required properties for a good performance. Various pavement materials viz., subgrade soil, Granular Sub-base (GSB), Wet Mix Macadam (WMM) base, Bituminous cores and chunk samples, collected from 11 pits, were subjected to detailed laboratory investigations.

70

A. Mittal and A. Kumar

4.1 Subgrade The subgrade soil was found to be poorly compacted in all the test pits, except pit number 3. The relative compaction values were found to be in the range varying from 91.5 to 99.3%, as against the specified value of 97% as per MoRT&H [4] specifications.

4.2 Granular Materials (GSB and WMM) The gradation of GSB was found to be meeting the specified gradation requirements, except for pit number 3, 4 and 8. However, the GSB layer was found to be poorly compacted as it was found to be not meeting the relative compaction requirements of 98% (except in pit no. 2) as per MoRT&H specifications. The gradation of WMM used in granular base was found to be meeting the specified requirements, except for pit number 2. However, the WMM layer was found to be poorly compacted as it was found to be not meeting the relative compaction requirements of 98% as per MoRT&H specifications.

4.3 Characteristics of Bituminous Mixes The samples of bituminous layers collected from the test pits and the core samples extracted were subjected to detailed laboratory testing. The combined cores extracted were sliced for top and lower bituminous layer and separated out in the laboratory, using a diamond cutter without disturbing their cylindrical shapes. The samples were tested for thickness, bulk density and percent bitumen. The extracted washed aggregates were also tested for various physical properties as per MoRT&H specifications. The density acceptance criterion given by MoRT&H for bituminous mixes is given as below:  Mean Density ≥ Specified density + 1.65 −

1.65 (No. of Samples)0.5



× Standard Deviation Here, the specified density is taken as 92% of the theoretical maximum specific gravity (Gmm ) (Tables 4 and 5). The binder content was found to be less than the specified requirements for all the core and chunk samples taken.

Forensic Investigations for Failure of Flexible Pavements: A Case Study

71

Table 4 Results for acceptance of density values Serial number Road type identification

Mean density Require density as per Remarks acceptance criterion

1

Samples of BC (main 2.196 carriageway 90 m roads)

2.291

2

Samples of BC (main 2.226 carriageway 45 m roads)

2.294

3

Samples of SDBC (Service road)

2.273

2.215

Not acceptable

Table 5 Results for acceptance of thickness values Serial number Road type identification

Mean thickness Required thickness as Remarks per design

1

Samples of BC 33.9 (main carriageway 90 m roads)

40

2

Samples of DBM 71.7 (main carriageway 90 m roads)

80

3

Samples of BC 32.6 (main carriageway 45 m roads)

40

4

Samples of DBM 63.8 (main carriageway 45 m roads)

80

5

Samples of SDBC (Service road)

28.8

25

Acceptable

6

Samples of DBM (Service road)

53.2

70

Not acceptable

Not acceptable

5 Probable Causes for Development of Distress The probable causes for development of distress/defects derived based on the experiences gained at field, observations made during the test pitting and coring as well as from the laboratory and field data/ results obtained through the investigation done on the project road, can be summarized as follows: (i) Overloading by commercial vehicles (trucks and multi-axles) as indicated by high VDF and number of commercial vehicles. (ii) Materials used in the various component layers not meeting the specified requirements at many locations.

72

A. Mittal and A. Kumar

(iii) Inadequate compaction of the pavement layers as indicated by the layers not meeting the density requirements, which has rendered the pavement prone to permanent deformation under the action of loaded commercial vehicles. (iv) Grossly inadequate structural capacity of the existing pavement and foundation for the traffic loads.

6 Proposed Short-Term and Long-Term Measures The recommendations have been proposed for immediate action (short-term measures) and strengthening through overlay (long-term measures).

6.1 Short-Term Measures (Till the Time Major Strengthening is Carried Out) • All existing surface defects like cracks, potholes, depressions/undulations/ deformations etc. shall be properly treated/filled up. • Proper sealing of cracks and filling up of undulations/depressions, wherever and whenever required, depending on the extent and severity of distress must be continued, as soon as they occur, on urgent/priority basis to minimise further progression of deterioration. • Stipulated routine and periodic maintenance to be carried out at regular intervals.

6.2 Long-Term Measures (Strengthening Through Overlay) The requirement of the flexible overlay to be provided for improving the structural adequacy of the existing roads has been worked out for 10 years design life, with the specification options as given below (Table 6). It is also recommended that the road stretch is recommended to be re-evaluated for any further structural and functional improvement requirements at the end of 5 years. This is recommended considering the fact as the industrial activity expands in the study area, the traffic is also expected to increase, both in terms of number and loading. So, to avoid any excessive damage to the roads in the area, it would be good to re-evaluate the roads for structural and functional improvement requirements.

Forensic Investigations for Failure of Flexible Pavements: A Case Study

73

Table 6 Overlay thickness proposed for design life of 10 years Serial number

Description of the road

Cumulative standard axles

Recommended overlay thickness

1

CETP plant to Mohna road 18.7 (Direction: towards Mohna road)

70 mm DBM + 50 mm SMA or 70 mm DBM + 50 mm BC

2

CETP plant to Mohna road 12.1 (Direction: towards CETP plant)

65 mm DBM + 50 mm SMA or 65 mm DBM + 50 mm BC

3

Mohna distributary to 18.7 entry at proposed bridge (Direction: towards Mohna distributary)

50 mm DBM + 40 mm SMA or 50 mm DBM + 40 mm BC

4

Mohna distributary to entry at proposed bridge (Direction: towards entry at proposed bridge)

12.1

50 mm DBM + 40 mm SMA or 50 mm DBM + 40 mm BC

5

45 m wide roads

18.7

75 mm DBM + 50 mm SMA or 75 mm DBM + 50 mm BC

Note SMA has been recommended as a better option being a stone aggregate strength derived, bituminous rich specification with a fibre content and better rut resistance. The fresh rehabilitation layers need to be produced using modified bitumen/VG-40 to derive the enhanced life of the pavement

References 1. CSIR-CRRI (2020) Investigation to study causes of failure of flexible pavements at IMT Faridabad and suggesting suitable remedial measures. New Delhi 2. IRC:81 (1997) Guidelines for strengthening of flexible road pavements using Benkelman beam deflection techniques. Indian Roads Congress, New Delhi, India 3. IRC:37 (2018) Guidelines for the design of flexible pavements. Indian Roads Congress, New Delhi 4. MORTH (2013) Specifications for road and bridge works. Ministry of Road Transport and Highways, New Delhi

Effect of Subgrade Stabilization on Pavement Design: Material Optimization and Economic Impacts Sudeshna Purkayastha, Ritu Raj Patel, Veena Venudharan, and Ajitkumar Vadakkoot

Abstract The objective of this research was to evaluate the effect of subgrade stabilization on the flexible pavement design. The stabilizer used for the study was naturally derived mineral stabilizer, making it a sustainable alternative to the currently employed soil stabilizers. The scope of this study included the evaluation of improved properties of stabilized subgrade through Proctor compaction, UCS, and CBR tests; followed by pavement design as per IRC 37: 2018; and then economic analysis. The experimental results indicated that an addition of 4% mineral stabilizer increased the CBR by 15 times. Further, an overall thickness reduction of ~30 and 40% was observed with 2 and 4% stabilization, respectively. The associated material optimization was in order of 45% for low and medium traffic levels, and 58% for high traffic levels at 4% stabilization. Economic analysis based on the construction cost of materials showed 23 and 30% reduction when soil was stabilized with 2 and 4% stabilization, respectively, as compared to untreated soil. Overall, this study illustrated the effect of subgrade stabilization with mineral stabilizer on the pavement design and its associated economic impacts. Keywords Subgrade stabilization · Pavement design · Economic analysis

S. Purkayastha · V. Venudharan (B) Indian Institute of Technology Palakkad, Palakkad, Kerala 678557, India e-mail: [email protected] S. Purkayastha e-mail: [email protected] R. R. Patel Pandit Deendayal Energy University, Gujarat 382355, India A. Vadakkoot Nisco Builders, Malappuram, Kerala 676504, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_8

75

76

S. Purkayastha et al.

1 Introduction The long-term performance of pavement structures is significantly impacted by the quality of pavement materials and stability of the different layers of the pavement structure. Among the various pavement materials, the in-situ subgrade soil plays a vital role in defining the pavement strength to withstand the traffic loads and environmental impacts. The subgrade’s structural properties not only decide the suitability of soil on pavement construction but also influence the subsequent pavement superstructure over the subgrade [1]. However, in the recent past, it has been observed that the unavailability of good subgrade is one of the major concerns in most of the pavement construction locations. Though previously the in-situ material was replaced by good soil from borrow area, it was not always practical and economically viable. Thus, the most commonly used methodology in improving the physical and mechanical properties of subgrade soil to meet the engineering purpose is soil stabilization [2]. Though, there are numerous methods of soil stabilization; use of additives like lime, cement, and asphalt are popular stabilization methods because of the ease in application, efficient, and economical [3]. But over the period, it was observed that multitude concerns followed these common stabilization methods. It was found that soil stabilized by cement or lime showed high pH content. A lot of research has been done on this domain, and it was observed that alkaline migration from stabilized soil depended upon the permeability of the soil and the surrounding cover soil [4]. Moreover, a lot of CO2 emissions due to cement production is seen, prevention of vegetation growth, groundwater contamination, and heat island creation, to name a few. Another drawback of calcium-based soil stabilizers includes the production of expansive products such as gypsum, ettringite, and thaumasite leading to cracking in soil. Such adverse influence from such stabilizers demands the need for green, sustainable, and effective nontraditional stabilizers, such as enzymes [5]. Substitution of conventional material (cement) and primary raw material (lime) with secondary raw material (waste and byproducts from industries) [6] corresponds to the Sustainable Development Goals set by the United Nations, preserves resources, saves energy, and reduces greenhouse gas emissions [7]. In this direction, mineral stabilization has received positive attention from the research community due to manifold benefits including, (i) naturally available, (ii) stabilization efficiency, (iii) recyclable, (iv) cost-effective, and (v) environmentally friendly. The mineral stabilizer also aids in [8]: . . . . .

Higher load-bearing capacity, tensile strength and improved modulus of elasticity Neutralization of pH levels and development of water-impermeable layers Excess heating or energy is not required Non-toxic and not harmful to health Recyclable up to 100%.

Effect of Subgrade Stabilization on Pavement Design: Material …

77

Thus, the main objective of the paper is to investigate the effect of a naturally developed mineral powder stabilizer on the structural properties of the soil subgrade and its effect on the pavement design. The scope of the study included: . Literature review pertinent to subgrade stabilization, and associated changes in pavement design and construction . Stabilization of subgrade soil with varying stabilizer dosages . Structural characterization of stabilized soil using laboratory tests . Design of pavements for various subgrade and traffic conditions . Economic analysis of all designed pavements. It is envisioned that this study will pave the way for the development of newer technologies on subgrade stabilization that are efficient, durable, and sustainable. The study will also aid in identifying the cost–benefit due to subgrade stabilization.

2 Materials and Experimental Program 2.1 Materials Soil. The subgrade soil used for the study was collected from a pavement construction site in the Alleppey region of Kerala (as shown in Fig. 1). The soil was laterite in nature, which is rich in iron and aluminium that are formed in tropical areas. It is important to note that lateritic soils in their natural state generally have a low bearing capacity and low strength due to high clay content, strength and stability in the presence of moisture cannot be guaranteed. The pavement construction at this site was impossible due to the poor strength properties of the soil subgrade and necessitated stabilization. An X-ray diffraction analysis (XRD) on the soil sample indicated that the soil mainly consisted of C, H, Na, O, S, Fe, and Al. The particle size analysis done as per IS 2720 (Part V) [9] indicated that the amount of particles passing 75 µm sieve is more than 55%. Further, the Atterberg’s limits determined in accordance to IS 2720 (Part V) [9] presented that the liquid limit, plastic limit, and plasticity index were, respectively, 63, 61, and 2%, representing a non-expansible highly plastic soil. Mineral Stabilizer. The mineral stabilizer is a commercially available soil stabilizing agent that is developed from 100% mineral components (as shown in Fig. 2). The stabilizer is a sustainable alternative to the currently utilized soil stabilizers. The XRD analysis on the mineral stabilizer indicated the stabilizer is a polycrystalline material with the presence of C, H, Cl, Na, and O.

78

S. Purkayastha et al.

Fig. 1 Soil sample

Fig. 2 Mineral stabilizer

2.2 Experimental Program For the purpose of determining the stabilizing capability of the mineral stabilizer, the following laboratory tests were carried out on stabilized subgrade at 0, 2, 4, and 6% of stabilizer dosages: . Proctor Compaction (in accordance with IS 2720: Part VII) . Unconfined Compressive Strength (in accordance with IS 2720: Part X) . California Bearing Ratio (in accordance with IS: 2720: Part XVI).

Effect of Subgrade Stabilization on Pavement Design: Material …

79

Based on the seed values obtained from the laboratory evaluation, pavement design for various levels of stabilization was determined using IITPAVE software in accordance with IRC 37: 2018 [10] for three traffic levels, i.e., 10, 75, and 150 MSA. Further, a comparative analysis was prepared to identify the possible material savings with subgrade stabilization and consecutively, the potential economic benefits. The results and analyses of the laboratory experiments, and pavement design comparisons are discussed next.

3 Results and Discussions 3.1 Laboratory Test Results Proctor Compaction. In order to determine the optimum moisture content (OMC) and maximum dry density (MDD), Standard Proctor test as per IS: 2720 (Part VII) [11] was conducted on the stabilized soil at various stabilizer dosages. Figure 3 presents the MDD results with the change in stabilizer dosage. As observed, the MDD increased with increase in dosage indicative of improved strength. The MDD of the natural soil subgrade was 1.415 g/cc at an OMC of 21%. With addition of the mineral stabilizer, a gradual increase in the MDD and reduction in OMC was observed. At 6% of dosage, the stabilized soil displayed MDD of 1.612 g/cc at an OMC of 17%. Unconfined Compressive Strength. UCS test is one of the fastest methods of measuring shear strength of the soil. UCS test in accordance to IS: 2720 (Part X) [12] was conducted in order to determine UCS of the soil subgrade and to identify the impact of stabilization on compressive strength. As shown in Fig. 4, with increase in stabilizer dosage, a significant increase in the UCS magnitude was observed. Note that the shear failures observed in the specimens were found to be symmetrical. While the untreated soil subgrade exhibited an UCS of 7.547 kg/cm2 , i.e., 740.104 kPa, with Fig. 3 MDD results of stabilized soil

1.65

MDD, g/cc

1.60 1.55 1.50 1.45 1.40

1.35 1.30 0

2 4 Stabilizer Dosage, %

6

80

S. Purkayastha et al.

Fig. 4 UCS results of stabilized soil

18 16

UCS, kPa

14 12

10 8 6 4 2

0

Fig. 5 CBR results of stabilized soil

0

2 4 Stabilizer Dosage, %

6

0

2 4 Stabilizer Dosage, %

6

70 60

CBR, %

50

40 30 20 10

0

mineral stabilization at 6% dosage, the UCS was found to be 1531.79 kPa. The UCS results indicated the improvement in compressive strength with soil stabilization. California Bearing Ratio. CBR is a measure of the strength of the pavement subgrade and is a vital input for the pavement design. CBR test was conducted in accordance to IS: 2720 (Part XVI) [13]. Figure 5 displays the soaked CBR results of the stabilized subgrade. The CBR of the natural subgrade was found to be 4% and with the inclusion of mineral stabilizer, the CBR of the soil was significantly enhanced. Note that the CBR was maximum with a magnitude of 64% at 4% of stabilization indicative of improved bearing strength. It is also important to note that the CBR value decreased with further increase in the stabilizer dosage. Therefore, for further analysis, stabilization up to 4% dosage is examined.

3.2 Pavement Design As an effort to identify the effect of subgrade stabilization on the pavement structure and its associated performance, a pavement design exercise was carried out in accordance with IRC 37: 2018. For the pavement design, the vital subgrade input property,

Effect of Subgrade Stabilization on Pavement Design: Material …

81

CBR was adopted from the laboratory results as mentioned in the previous section. Further to evaluate the interaction effect of both subgrade stabilization and traffic levels on the pavement design, three traffic levels, 10, 75 and 150 MSA were considered in the study to represent the extreme and also the intermediate traffic conditions. The pavement design was accomplished using IITPAVE software available with IRC 37: 2018. Figure 6a and b represents the pavement cross-sections considered for various traffic levels. As shown in Fig. 6a, the cross-section for 10 MSA traffic level includes, granular subbase (GSB), wet mix macadam (WMM), surface course with VG40 binder. In case of traffic levels, 75 and 150 MSA, in addition to GSB, WMM and surface course; a layer of dense bituminous macadam (DMB) with VG40 binder is considered as presented in Fig. 6b. Note that all layers were in accordance with MoRTH specifications (MoRTH 2013) [14]. Table 1 presents the pavement cross-sections for various traffic levels. Note that the pavement sections were designed by comparing the vertical strain at the top of subgrade (Ev ) and tensile strain at the bottom of bituminous layer (Et ) from the traffic and pavement structure. A safe pavement cross-section is the one with lower Ev and Et than the derived critical strains from the traffic conditions. Table 1 provides the cross-sectional details of pavements for different traffic and stabilization levels. From Table 1, it was found the pavement composition altered significantly with the inclusion of mineral stabilizer in subgrade, with a reduction in total pavement thickness with increase in the stabilizer dosage. It resulted in satisfactory values for Ev and Et at lower pavement layer thicknesses. Note that with 4% dosage of stabilization, Surface Course Wet Mix Macadam

Granular Sub-base

(a)

Subgrade Surface Course Wet Mix Macadam

Dense Bituminous Macadam

Granular Sub-base

(b)

Subgrade

Fig. 6 Pavement cross-sections for traffic of a 10 MSA and b 75 and 150 MSA

82

S. Purkayastha et al.

Table 1 Pavement cross-sections for various traffic levels Traffic level, MSA 10

75

150

Stabilizer dosage (%)

Pavement cross-section, mm Surface course

DBM

WMM

GSB

0

40

–

250

300

2

40

–

150

200

4

40

–

100

200

0

40

80

250

300

2

40

80

150

200

4

40

80

100

200

0

40

100

350

400

2

40

80

120

200

4

40

80

110

200

the overall pavement thickness reduced approximately by 45% for low and medium traffic levels and the reduction was more than 50% for high traffic level pavements. Similarly, it was also observed that an increase in traffic level necessitated a thicker pavement cross-section. Higher the pavement layer thicknesses, lower the Ev and Et , and therefore, safe pavement design for higher traffic levels. With an increase in traffic level from 10 to 150 MSA, the overall pavement thickness increased approximately by 1.5 times.

3.3 Economic Analysis This sub-section discusses the reduction in construction material required and associated savings in construction cost. In order to carry out the aforementioned tasks, a pavement of length 1 km and 3.5 m width was studied, Table 2 presents the material requirement and construction cost for all the nine designed pavement crosssections. As observed, the construction material requirement drastically reduced with the inclusion of mineral stabilizer in the subgrade. For a traffic level of 10 MSA, with the addition of 2% stabilizer to the subgrade, the required construction material reduced by 36%, wherein the reduction in WMM was approximately 40% and in GSB was nearly 33%. Similar observations were made in case of higher traffic levels as well. With 2% stabilizer dosage, ~36% material reduction was observed in case of 75 MSA traffic and ~57% material reduction were observed in case of 150 MSA traffic. Furthermore, with 4% dosage of subgrade stabilization, nearly 45% reduction in construction materials was noticed for traffic levels, 10 and 75 MSA; and 58% reduction was observed in case of 150 MSA traffic. It is concluded that subgrade stabilization not only improves the subgrade strength but also significantly lowers the material need for the pavement construction.

Effect of Subgrade Stabilization on Pavement Design: Material …

83

Table 2 Material requirement and economic analysis of pavement cross-sections Traffic level, MSA 10

75

150

Stabilizer dosage (%)

Construction material (m3 ) Surface course

DBM

WMM

GSB

Construction cost (|)

Savings in construction cost (%)

0

140

0

875

1050

3,945,641

–

2

140

0

525

700

2,932,391

25.7

4

140

0

350

700

2,672,866

32.3

0

140

280

875

1050

4,729,641

–

2

140

280

525

700

3,716,391

21.4

4

140

280

350

700

3,456,866

26.9

0

140

350

1225

1400

5,938,891

–

2

140

280

420

700

3,560,676

40.0

4

140

280

385

700

3,508,771

40.9

Additionally, the construction cost analysis carried out on all nine pavement crosssections displayed the cost savings due to subgrade stabilization. Subgrade stabilization averted the need of thicker pavement cross-section and in turn resulted in significant decrement of pavement construction cost. As presented in Table 2, with a stabilizer dosage of 2% in subgrade, around 25% savings in construction cost was observed for lower traffic level of 10 MSA, and around 21% savings was observed for medium traffic level of 75 MSA. For high traffic level of 150 MSA, a rise in overall savings was observed with nearly 40% reduction in construction cost. Similar observations were made for higher level of subgrade stabilization. With increase stabilization dosage by 2%, an increment in total savings was observed for all traffic levels. The increments were in orders of approximately 7, 5, and 1% for 10, 75 and 150 MSA traffic. Overall, it was concluded that the subgrade stabilization resulted in thinner pavement cross-sections and lesser construction cost in comparison with the un-stabilized subgrade.

4 Conclusion and Future Scope This research work evaluated the performance and economic benefits of subgrade stabilization using a newly available mineral stabilizer. The properties of the stabilized soil were evaluated through essential tests, including Proctor Compaction, Unconfined Compressive Strength, and California Bearing Ratio; and the results were considered for flexible pavement design as per IRC37:2018. The major findings were summarized as follows: . Proctor Compaction: MDD was found to be increasing with an increase in the percentage of mineral stabilizer. An increment of 15% in MDD was observed when the subgrade was stabilized with 6% mineral stabilizer. Further, it was also

84

. .

.

.

S. Purkayastha et al.

observed that with the inclusion of stabilizer the amount of water required was reduced significantly. UCS: was also found to be increasing with increase in the stabilizer dosage. With 6% stabilizer dosage, the UCS value increased by nearly 100%. The UCS results signified a definite improvement in the soil’s compressive strength. CBR: showed substantial improvement with an increase in mineral stabilizer content. With the addition of 4% stabilizer, the CBR value increased by 15 times. Higher CBR subgrade means a comparatively thinner subbase layer is required to satisfy the traffic load requirement of the pavement. However, it was also observed that with further increase in the stabilizer dosage, the CBR decreased indicative of an optimum stabilizer dosage. Pavement Design: was done for different dosages of mineral stabilizer to determine the thickness of various layers for three traffic levels. 30 and 40% of overall thickness reduction were observed with the addition of 2 and 4% mineral stabilizer, respectively. Economic Analysis: was done for different dosages of mineral stabilizer and for low and medium traffic levels, a general decrease in thickness resulted in nearly 23 and 30% cost saving when the soil was stabilized with 2 and 4% stabilization, respectively. The cost savings were the highest for 150 MSA traffic level.

The future scope of the study includes more research towards exploring the influence of the stabilizer on other soil types and its associated variations in stabilization, micro-characterization, and optimization. In this direction, it is also important to understand the permeability properties of the soil and the impact of stabilization on subgrade permeability and drainage characteristics. Further, it is also planned to conduct a comparative analysis on effect of subgrade stabilization on both rigid pavement and flexible pavement performance and its life cycle analysis. Overall, it is envisioned that this study will help understand the performance and economic benefits of stabilized subgrade using a sustainable alternative approach.

References 1. Estabragh AR, Jandari F, Javadi AA, Amini M (2022) Effect of magnesia on stabilization of contaminated clay soil. ACI Mater J 119(3):103–113 2. James J, Pandian PK (2013) Performance study on soil stabilisation using natural materials. Int J Earth Sci Eng 6(1):194–203 3. Ramachandran AL, Dubey AA, Dhami NK, Mukherjee A (2021) Multiscale study of soil stabilization using bacterial biopolymers. J Geotech Geoenviron Eng 147(8):04021074 4. Renjith R, Robert DJ, Gunasekara C, Setunge S, O’Donnell B (2020) Optimization of enzymebased soil stabilization. J Mater Civ Eng 32(5):04020091 5. Vincevica-Gaile Z, Teppand T, Kriipsalu M, Krievans M, Jani Y, Klavins M, et al (2021) Towards sustainable soil stabilization in peatlands: secondary raw materials as an alternative. Sustainability 13(12): 6726 6. Soldo A, Mileti´c M, Auad ML (2020) Biopolymers as a sustainable solution for the enhancement of soil mechanical properties. Sci Rep 10(1):1–13 7. Optimization of enzyme-based soil stabilization

Effect of Subgrade Stabilization on Pavement Design: Material …

85

8. Kavak A, Coruk Ö, Aydıner A. A new binder mineral for cement stabilized road pavement soils 9. Guidelines for the design of flexible pavements Indian Road Congress, IRC (2018) fourth revision. Indian Road congress, New Delhi 10. IS: 2720 (Part V) (1972) Determination of liquid and plastic limit, Indian Standard Code. Indian Standard Institution, New Delhi 11. IS: 2720 (Part VII) (1983) Determination of water content-dry density relation using light compaction. Indian Standard Institution, New Delhi 12. IS: 2720 (Part X) (1991) Determination of unconfined compressive strength, Indian Standard Code. Indian Standard Institution, New Delhi 13. IS: 2720 (Part XVI) Determination of California bearing ratio, Indian Standard Code (1991). Indian Standard Institution, New Delhi 14. Specifications for road and bridge works, MORTH (5th revision) (2013) Indian Road Congress, New Delhi

An Effective Bitumen-Friendly Polymer for Superior Roadway Performance and Durability Krishna Srinivasan, Sachin Raje, and Deepak Madan

Abstract Polymer Modified Bitumens (PMB) are utilized in Asphalt Mixes to improve high temperature grade, stiffness, rutting resistance and damage tolerance, resulting in more durable roadways which are able to withstand increased traffic loads. PGXpand, a Bitumen-Friendly Polymer (BF Polymer), developed by Sripath Technologies, is unlike any traditional elastomeric or plastomeric polymers. The BF Polymer interacts with the binder in a very unique fashion, imparting key performance benefits to PMB Mixes, while mitigating the processing difficulties and shortcomings typically associated with traditional polymers. Increasingly, experts from around the globe believe that performance and durability of roadways is more important than measuring elastic recovery of the modified binder, which only confirms the presence of a polymer [1, 2]. This is where the BF Polymer shines. It is designed to interact with bitumen in a very innovative and unique manner. The BF Polymer is designed to be highly dosage efficient, storage stable, and easy to mix into bitumen. It is designed to improve the workability of PMB Mixes and mimic the advantages of warm mix additives. It is designed to lower the paving temperatures and make the mix easier to pave and compact. The BF Polymer is designed to deliver the desired roadway performance and durability for targeted weather and traffic conditions. Roadways paved using BF Polymer PMBs deliver outstanding rutting resistance, excellent fatigue properties, and long-term durability.

K. Srinivasan · D. Madan (B) Sripath Technologies, LLC, Mahwah, NJ, USA e-mail: [email protected] K. Srinivasan e-mail: [email protected] S. Raje Bitpath Pvt. Ltd., Mumbai, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_9

87

88

K. Srinivasan et al.

Keywords Bitumen-Friendly Polymer · PGXpand polymer · Polymer modified bitumen (PMB)

1 Introduction PGXpand, a Bitumen-Friendly Polymer (BF Polymer), developed by Sripath Technologies, is a novel, medium molecular weight polymer, specifically engineered with a tailored architecture and chemistry [1]. It is a new family of polymer additives, unlike any traditional elastomeric or plastomeric polymers. The BF Polymer interacts with the binder in a unique fashion. It delivers certain key performance benefits to PMB Mixes, while making it easier to process the PMB by mitigating the difficulties and shortcomings of traditional polymers.

1.1 Key Characteristics of Bitumen-Friendly Polymer Additive Figure 1 summarizes key characteristics of the BF Polymer presented in this paper. BF Polymer delivers excellent performance, it is easy to incorporate into a plant operation, and it helps reduce the overall cost of the mix. When compared to a PMB made using traditional polymers, a PMB produced with the BF Polymer is much easier to incorporate into a Mix plant. It melts rapidly in Bitumen at around 150 °C. It can be easily incorporated into the binder using low energy mixers and short mixing times. It exhibits excellent storage stability. The BF Polymer is highly dosage efficient. At low dosage levels, it lowers the viscosity of bitumen and enhances internal lubricity properties in Mixes. It improves the workability of PMB Mixes and mimics the advantages of a warm mix additive. It lowers the paving temperature and makes the mix much easier to compact.

Fig. 1 Key characteristics of BF polymer

An Effective Bitumen-Friendly Polymer for Superior Roadway …

89

The BF Polymer eliminates or reduces the need for raw materials such as, crosslinking additives, warm mix additives, and traditional polymers. It contributes to lower energy consumption and overall lower mix and paving costs. This polymer boosts the high temperature performance of bitumen without any impact on the low temperature properties. Roadways paved with PMB mixes made using BF Polymer deliver outstanding rutting resistance, excellent fatigue properties, and long-term durability.

1.2 Key Materials Used and Test Standards Binders, aggregates, and other asphalt additives from around the globe were used when generating the data presented in this paper. The BF Polymer used for the data in this report was PGXpand, a Bitumen-Friendly Polymer developed and commercialized by Sripath Technologies [1, 2]. BF Polymer has been compared against commercially available traditional plastomeric and elastomeric polymers. Test data presented was conducted using materials and as per specifications applicable to the geographic region where testing was conducted.

2 Results and Discussions The BF Polymer has been tested, evaluated, and vetted by leading experts in academia, industry and transportation agencies from around the world. It has been trusted and effectively used on roadways across the globe since 2015 [2].

2.1 Improved High Temperature Properties The impact of varying the BF Polymer content on the properties of a commercial grade of bitumen grade PG 64-22, sourced in the USA was evaluated. Table 1 outlines the impact of the addition of 1.5 to 3% of the BF Polymer to this bitumen. The BF Polymer raises the softening point of the base binder from 48 °C to over 120 °C for a 2% polymer addition. Penetration values are reduced from 67 dmm to around 40 dmm and the viscosity reduces from 450 cps to around 380 cps with the addition of the BF Polymer. The viscosity profiles shown in Fig. 2 compare the viscosity versus temperature of the base bitumen (USA Grade 64-22) and the base bitumen modified with 1.5% of the BF Polymer. The addition of the polymer lowers the viscosity of the binder by 5–10% over the entire production-compaction temperature range, while boosting the high temperature grade.

90

K. Srinivasan et al.

Table 1 Hot mix properties of PMB mix USA bitumen PG 64-22 + Bitumen-Friendly Polymer; as per applicable ASTM standards With PGXpand®

Softening point (°C)

Penetration (dmm)

True grade (°C)

Viscosity @ 163 °C (cps)

Base binder

48

67

66.8–26.1

450

1.5%

98

45

85.3–25.1

402

2%

122

40

95.0–25.0

384

3%

123

39

95.0–25.0

370

Fig. 2 Viscosity profile for USA base bitumen versus binder + BF polymer

The Royal Melbourne Institute of Technology in Melbourne, Australia (RMIT), evaluated the impact of the BF Polymer on the properties of an Australian bitumen C170 [3, 4]. Binder and mix level properties were evaluated on an AC14 dense grade mix. Both results are summarized in Table 2. Table 2 Properties of Australian bitumen C170 + Bitumen-Friendly Polymer AC14 dense graded asphalt mix, evaluated at RMIT, Australia Properties

Test method

Units

% PGXpand addition 1.2

1.5

2.0

2.5

3.0

Viscosity @ 165 °C

AS/NZS2341.4

Pa.s

0.11

0.11

0.12

0.11

0.12

Softening point

AGPT/T131

°C

74

107

111

115

117

Stress ratio @ 10 °C

AGPT/T125

−

0.99

1.02

1.05

1.02

1.03

Stiffness @ 25 °C

AGPT/T121

kPa

35

40

48

56

64

PMB separation

ASTM D7173

°C

1.1

0.8

0.2

1.1

1.5

Elastic Rec. @ 25 °C

ASTM D6084

%

22.0

19.0

13.5

13.0

11.0

Hamburg wet, rut depth @ 50 °C, 20 K

AASHTO T324

mm

4.88

4.62

3.57

2.70

2.18

An Effective Bitumen-Friendly Polymer for Superior Roadway …

91

Table 3 Properties of Indian bitumen VG30 + Bitumen-Friendly Polymer 5.4% binder, 3.84% voids, 12 mm, evaluated at IIT Bombay Properties

Test method

Units

1.5% PGXpand

Viscosity @ 150 °C

ASTM D4402

mPa.s

244.4

Softening point

IS 1205

°C

89.7

Penetration

IS 1203

dmm

20

PMB separation

IRC SP 53:2010-A3

°C

0.90

Elastic Rec. @ 25 °C

IRC SP 53:2010-A2

%

20.67

Hamburg wet, rut depth @ 50 °C, 20 K

AASTHOT 324

mm

5.69

Ideal-CT @ 25 °C, 50 mm/min

ASTM D8225-2019

−

74

Resilient modulus @ 1 Hz, 35 °C

ASTM D4123

MPa

2181

A typical dosage level of 1.5 to 2.5% of the polymer delivered excellent properties. The softening point increased to around 110 °C, stiffness to around 50 kPa, while maintaining low levels of viscosity of around 0.1 Pa.s when tested at 165 °C. The rut depth as measured by Hamburg Wet Wheel Tracking tested at 50 °C was around 4 mm for 20,000 cycles. A PMB based on Indian bitumen grade VG30 with 1.5% BF Polymer was evaluated at the Indian Institute of Technology Bombay, Mumbai, India (IIT Bombay) [5]. Mix level properties were evaluated on a mix with 5.4% binder, 3.84% voids and based on a nominal 12 mm aggregate. The properties are summarized in Table 3. The PMB with 1.5% of the BF Polymer delivered excellent properties—a viscosity of 244 mPa.s at 150 °C, a softening point of 90 °C, penetration at 20 dmm, and PMB separation of 0.9 °C. The rut dept as per Hamburg Wet Wheel Tracking tested at 50 °C was 5.7 mm and the Ideal CT Index tested at 25 °C was 74.

2.2 Superior Stability and Separation Properties PMB based on the BF Polymer exhibited excellent storage stability. Bitumen samples with 1.25% of the polymer were held in a cigar tube at 165 °C for 72 h. G*/Sinδ of the top and bottom portions of the cigar tube were measured at 3 different temperatures. The variation between the top and bottom measurements was less than 5%. Other properties, such as softening point, penetration, and phase angle, showed similar small variations, demonstrating the outstanding storage stability of the BF Polymer. Also, as seen in Table 2, an Australian Bitumen C170 mixed with BF Polymer at levels ranging from 1.2 to 3.0% showed excellent PMB separation properties.

92

K. Srinivasan et al.

Fig. 3 Comparing compactability curve of BF polymer PMB versus no polymer hot mix Australian bitumen C170 + BF polymer versus hot mix based on C320 bitumen

2.3 Unique Mix Compaction and Workability Even at low dosage levels, the BF Polymer lowers the viscosity of bitumen and enhances its internal lubricity properties, resulting in improved workability of such PMB Mixes. The polymer mimics the advantages of a warm mix additive, lowers the paving temperature and makes the mix much easier to pave and compact. Figure 3 compares the compactability curves of an Australian AC14 dense-graded Mix based on a C170 Australian binder and 3% BF Polymer to the curve for a standard Hot Mix Asphalt (HMA), made using a C320 Australian bitumen with no polymer added. The two curves are almost identical, demonstrating the unique ability of the BF Polymer to improve Mix workability and aid compaction. It should be noted that the softening point of the bitumen in the BF Polymer based mix is over 110 °C compared to the softening point of around 50 °C for the standard HMA.

2.4 Excellent Rutting Resistance An in-depth study on the rutting performance of the BF Polymer was conducted by the Indian Institute of Technology Madras, Chennai, India (IIT Madras) [6]. Figure 4 compares the rutting resistance, as measured by a Hamburg Wheel Tracking Test at 60 °C, of an Indian bitumen grade VG30 to two PMB mixes, one made with 3.5% traditional elastomeric polymer and the other with 1.2% BF Polymer [6]. The bitumen in both PMB mixes had a softening point of around 75 °C. The PMB Mix made with BF Polymer has far superior rutting resistance. By any measure of rutting resistance, the BF Polymer consistently showed superior rutting resistance compared to the traditional elastomeric polymer.

An Effective Bitumen-Friendly Polymer for Superior Roadway …

93

Fig. 4 Rutting resistance of Indian bitumen VG30 mixes base binder versus 3.5% SBS PMB versus 1.25% BF polymer PMB Hamburg Wheel Tracking Test @ 60 °C, Indian bitumen VG30

2.5 Fatigue Performance In addition to providing superior high temperature performance and rutting resistance, PMB mixes made using the BF Polymer, in spite of low elastic recovery values, demonstrated fatigue performance comparable to traditional elastomeric PMB mixes. The 4-point beam bending fatigue performance of two PMB mixes is compared in Fig. 5. The curve for the PMB mix made using 1.75% BF Polymer is almost identical to the curve for the PMB mix made using 4% traditional elastomeric polymer. The flexural fatigue test was conducted at 20 °C, 10 Hz loading, using a PG 64-22 binder grade for a high-performance bridge deck mix in New Jersey, USA. Both the BF Polymer and the traditional elastomeric modified bitumens had a high temperature true grade of 75 °C. Fig. 5 Fatigue 4-point beam bending @ 20 °C, 10 Hz USA bitumen PG 64-22, 4% SBS PBM versus 1.75% BF polymer

94

K. Srinivasan et al.

Fig. 6 Bitumen-Friendly Polymer on expressway in India

3 Key Applications of Bitumen-Friendly Polymers Bitumen modified with BF Polymer is used for a variety of paving and road repair and maintenance applications, including: hot mix paving, hybrid polymer modified bitumen, high-stiffness asphalt mixes (HiSAM), hot spray sealing, BF Polymer modified emulsions, and cold patch mixes for pothole repair.

3.1 Hot Mix Paving Applications Polymer Modified Bitumen (PMB) manufactured using BF Polymer has been used to improve rutting resistance and deliver desired roadway performance on high-traffic roadways around the world since 2015. Such PMBs are easy to produce, help reduce cost and are easier to pave and compact. As an example, the BF Polymer was successfully used to pave a wear layer surface on a 2 km stretch of a high traffic state highway SH-2 in Andhra Pradesh, India [7]. About 1.5% of the BF Polymer mixed with an Indian bitumen VG-30 yielded properties similar to a mix based on 3.5% of a traditional elastomeric polymer (Fig. 6). As another example, the BF Polymer was successfully used to improve the rutting performance on a 4-lane km stretch of Agra-Gwalior Highway in India [8]. About 1.5 wt.% of BF Polymer was added to an Indian bitumen grade VG30. The Mix based on this PMB was used to pave a 30 mm wear layer on a 4 lane-km stretch of Agra-Gwalior highway about 40 km from Gwalior.

3.2 Hybrid Polymer Modified Bitumen As seen in Table 4, the BF Polymer can be used in combination with traditional elastomeric polymers to create Hybrid PMB formulations. Table 4 shows the effect of adding BF Polymer to a PMB created by adding 2.5% traditional elastomeric polymer to the base binder. An addition of 1.5 to 3% of BF Polymer to the PMB,

An Effective Bitumen-Friendly Polymer for Superior Roadway …

95

Table 4 Hybrid polymer modified bitumen properties of USA bitumen PG 64-22 + 2.5% SBS + BF polymer With PGXpand Base binder + 2.5% SBS

Softening point (°C) 66

Penetration (dmm)

True grade (°C)

Viscosity @ 163 °C (cps)

74

73.1–29.0

450

1.5%

105

47

85.6–28.4

402

2%

125

42

95.0–28.1

384

3%

124

38

95.0–28.0

370

increased the softening point, lowered the penetration, and reduced the viscosity by about 7–10%. Such a Hybrid PMB allows the incorporation of select benefits of BF Polymer into a traditional elastomeric PMB, making them easier to produce and use, deliver improved roadway performance.

3.3 High Stiffness Asphalt Mixes (HiSAM) Currently available high-stiffness mixes are quite expensive, need to be paved at high temperatures, require multiple lifts to lay down the road, are difficult to manufacture and are not easy to use in the field. High-Stiffness Asphalt Mixes based on BF Polymer (HiSAM) are a viable alternate solution. HiSAM mixes are easy to manufacture and use, allow paving at lower temperature, and are environmentally more sustainable. The technology allows for the reduction in void content, reduction in base course thickness, improvement of moisture resistance, and consolidation of one or more lifts.

3.4 Hot Spray Seal Applications BF Bitumen is also used in hot spray chip seal applications. The BF Bitumen can be easily incorporated into bitumen, in some cases along with crumb rubber or traditional elastomeric polymers. Such PMBs have a lower viscosity, are easy to spray, result in better roadway performance, and potentially help reduce cost.

96

K. Srinivasan et al.

3.5 PGXpand Modified Emulsion Applications The BF Polymer can be easily incorporated into bitumen to create a PGXpand Modified Emulsion. Such emulsions are easy to emulsify using traditional emulsification technologies and applied on a road surface using normal application techniques, equipment and practices. BF Polymer based emulsions are used for slurry seal, chip seal, scrub seal, fog seal, and microsurfacing applications. The BF Polymer promotes rapid skin-over, resulting in a tack-free trackless road surface that can be rapidly opened to traffic. Key advantages include, less greenhouse gas emissions, improved emulsion stability for transport, outstanding rutting resistance, and durable roadways that require less frequent maintenance.

4 Ease of Use in a Plant Operation The BF Polymer is easy to incorporate into bitumen in a PMB plant, a mix plant. Table 5 outlines the recommended guidelines for incorporating BG Polymer into a Mix plant operation [9]. The addition of the BF Polymer into bitumen is an endothermic process, thus, the mixture needs to be heated during the entire blending process to maintain the desired temperature. It is important to note that when BF Polymer is added to bitumen, the viscosity drops by about 5 to 15%. At suggested blending temperatures, the BF Polymer melts easily into the bitumen, due to liquidon-liquid mixing. At a blending temperature of about 150 to 160 °C, the BF Polymer can easily be mixed into bitumen within 1 to 2 h using low shear mixers, and minimal recirculation. Table 5 Ease of using Bitumen-Friendly Polymer Key parameters

Guidelines

Blending temperature

Endothermic reaction 150 to 163 °C (300 to 325 °F)

Mixing equipment

Low shear mixer

Blending speed

Maintain a small minimal vortex

Blending time

1 to 3 h

Paddle speed and temperature

Reasonable speed above 163 °C (325 °F)

Viscosity

Drops with addition of PGXpand

An Effective Bitumen-Friendly Polymer for Superior Roadway …

97

Fig. 7 Bitumen-Friendly Polymer reduces overall mix cost

5 Reducing Overall Mix Cost with Bitumen-Friendly Polymer The use of BF Polymer yields important cost benefits. Paving mixes typically require less than half the amount of a traditional polymer to yield similar properties. It is typically possible to replace traditional elastomers in a 2:1 or better ratio to yield similar roadway performance. This reduction in raw materials requirement results in overall savings in mix cost, as demonstrated in Fig. 7.

6 Conclusions The Bitumen-Friendly Polymer has been specially engineered to interact with bitumen in a very unique and innovative manner, allowing product to deliver superior roadway performance and durability. The polymer improves key performance qualities of bitumen, such as, softening point, penetration, and true high grade. It lowers viscosity of PMB mixes, improves workability and makes mixes easier to compact. This unique polymer delivers roadways with outstanding rutting resistance, fatigue properties and long-term durability. The BF Polymer has been tested, evaluated, and vetted by leading experts in academia, industry and transportation agencies from around the world. It has been trusted and effectively used on roadways across the globe since 2015. In the future, the authors plan to continue evaluating and reporting on the performance of adoption of BF Polymer in PMB mixes on roadways around the world. Acknowledgements The authors acknowledge the dedication of Bitpath Pvt. Ltd. in representing Sripath products, https://sripath.com/, in India and providing outstanding service to customers. Further, the authors are thankful for the invaluable technical guidance provide by experts from around the world.

98

K. Srinivasan et al.

References 1. Sripath Technologies Website. http://www.sripath.com. Accessed 2022/07/10 2. PGXpand® A unique polymer for paving and roofing applications, Sripath Technologies. https:// sripath.com/resources/documents/. Accessed 2022/07/10 3. Giustozzi F (2022) Project report on PGXpand by Royal Melbourne Institute of Technology, Melbourne, Australia, for Sripath Asia-Pac Pty. Ltd. 4. Hamid T (2022) Sripath PGXpand, a specially-engineered innovative polymer, vetted and evaluated by RMIT. In: Roads & Infrastructure 5. Singh D, Suchismita A, Krishna V (2022) Performance evaluation of PGXpand, Progress report by Indian Institute of Technology Bombay, Mumbai, India for BitpathPvt. Ltd. 6. Project Report (2021) Characterization of rutting performance of bituminous mixtures with PGXpand, by Indian Institute of Technology Madras, Chennai for Bitpath Pvt. Ltd. 7. Madan D (2021) PGXpand makes way for PMB in India’s Highway Mixes. In: AsphaltPro, pp 22–23 8. Sripath PGXpand improves rutting of major highway in India. https://sripath.com/sripath-pgx pand-improves-rutting-of-major-highway-in-india/. Accessed 2022/07/10 9. PGXpand® guidelines for blending the polymer into bitumen, Sripath Technologies. https://sri path.com/resources/documents/. Accessed 2022/07/10

Rheological Investigation of Soft Grade Asphalt Binder Modified With Crumb Rubber-Nanosilica Composite Tabish Mehraj, Mohammad Shafi Mir, and Bijayananda Mohanty

Abstract This study aims at investigating the effect of using nanosilica (NS) as a modifier for Crumb rubber (CR) modified asphalt binder. In this study, the concentration of CR was kept constant as 12% (wt of base binder) and the nanosilica concentration was varied from 1% to 6%. The effect of varying concentrations (1, 2, 3, 4, 4.5, 5 and 6%) of nanosilica (by weight of binder) on CR modified binder were evaluated by utilizing various physical tests like penetration, softening point, and ductility. The rotational viscosity (RV) and dynamic shear rheometer (DSR) tests were used to analyze rheological properties of base binder and nanosilica polymer modified asphalt binder. In addition, the performance of modified asphalt after thin film oven (TFO) (short-term aging) and Pressure aging Vessel (PAV) test (long term aging) were assessed as well. Furthermore, the storage stability of modified asphalt binder was evaluated. Results showed that the addition of nanosilica has a positive effect on rutting performance of CR modified asphalt binders. Storage stability of the CR modified asphalt binders improved significantly after the addition of nanosilica. Using softening point and rheological parameters (complex modulus (G*) and phase angle (δ), the best values were possessed by 12% CR-4% NS modified binder. During rheological characterization, it was found that complex modulus increases, phase angle decreases, superpave rutting parameter increases and failure temperature increases with increasing nanosilica content. It was also found that Brookfield viscosity increases with increasing nanosilica concentration as the binder becomes stiffer. All the test results confirmed the fact that the crumb rubber-nanosilica modifier is effective in enhancing the high temperature properties (rutting resistance) of the soft grade binders and at the same time, it increases the elasticity of the binders. Keywords Crumb rubber · Nanosilica · Superpave rutting parameter · Superpave fatigue parameter · Viscosity · Ageing · Storage stability

T. Mehraj (B) · B. Mohanty Department of Civil Engineering, National Institute of Technology, Mizoram 796012, India e-mail: [email protected] M. S. Mir Department of Civil Engineering, National Institute of Technology, Srinagar 190006, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_10

99

100

T. Mehraj et al.

1 Introduction Today is a world of sustainable technology and researchers are interested in finding sustainable, eco-friendly and cost-effective materials. Use of crumb rubber, also called black pollutant [34] in asphalt production is such cost-effective, ecofriendly [4, 30] and sustainable material which transforms an unwanted residue into a new bituminous mixture which is highly resistant to rutting and fatigue. The use of crumb tire rubber (CTR) is an interesting alternative from both economical and environmental perspectives [13]. CTR is dangerous due to its potential environmental threat and fire hazards [17] and its usage in pavements solve its disposal problem [37] and such threats. Crumb rubber modified binder (CRMB) provides improved mechanical properties, increases pavement durability and reduces reflective cracking and fatigue resistance [23]. Compared to conventional road surfaces, those made from CRMB have a longer service life, with 50–70% decrease in noise level raised cold- and heat-resistance properties and improved slip resistance, resulting in shorter braking distance and a higher safety coefficient [34]. Crumb rubber can be incorporated into asphalt mixtures either by wet or dry process. In the dry process, CR is used as partial substitution for fine aggregate. However in wet process, the CR powder is first blended with hot base asphalt and then swelled in the matrix to prepare modified asphalt. The wet process can generate two totally different modified asphalts, called as asphalt rubber (AR or wet-processhigh-viscosity) and terminal blend (TB or wet-process-no-agitation). AR binders can be produced by using more coarser rubber (minus 30 mesh, greater than 15% wt of virgin asphalt), and are resistant to rutting and reflective cracking but cause issues due to higher viscosity such as difficult to handle while paving, inability to store over long periods and modification of mixing equipment by contractors [15]. Thus, a good alternative is the TB binder that overcomes the viscosity issue by using less and finer CR (plus 30 mesh, less than 10%wt of virgin asphalt) and by applying high temperature shear in the modification. TB has better storage stability than AR [24]. TB binder holds many advantages such as low viscosity, good workability and applicability to dense graded mixture [25] but degradation of CR might impair the elasticity of binder and high temperature performance [22, 24]. Although digestion of rubber in TB improves its thermal storage stability it still has some separation problems due to partially undissolved rubber particles [14]. To remove such anomaly, modification with a second modifier can be done. In recent years, nanomaterials have been widely used to enhance the physical, rheological and mechanical properties of asphalt binder [29]. Addition of nano-sized additives improves the performance of asphalt binder and overcomes the drawbacks of polymers. Thus, Polymer nanocomposites are considered to be more powerful modifiers [28]. Nanosilica is an inorganic nanomaterial that has been being widely used to enhance the performance of polymer modified asphalt [1]. Nanosilica can be produced by agricultural waste materials such as rice husk ash, sorghum vulgare seed heads and bagasse ash by precipitation method, bio-digestion process and sol–gel process (Bhat and Mir 2020). Addition of Nanosilica improves the storage modulus,

Rheological Investigation of Soft Grade Asphalt Binder Modified With …

101

elasticity and ageing resistance of the asphalt binder [27] because of its high specific surface area, high functional density and high strain resistance. Various researches have been done to improve the physical, rheological, and mechanical properties of CRMA in recent years. For example, Attia has added 2% SBS to improve the binder performance [6]. Khasawneh added MSW and NS to CRMB and found no enhancement by MSW of high temperature performance but a negative effect on low-temperature performance. Addition of 12% CTR was recommended because it produced the greatest enhancement for high temperature performance [5]. Abedali added bentonite (khawa clay) and crumb rubber to asphalt. The addition of bentonite and crumb rubber enhanced asphalt properties such as the viscosity, ductility and softening point, decreased penetration [2]. Liu added NS to pre-oxidized CR. H2 O2 and NaClO were used to oxidize the CR. Overall, Storage stability and high temperature performance was enhanced [18]. The purpose of the current study is to use nanosilica to improve the performance of CRMB at high and intermediate temperatures as well as its storage stability. The goal of this study is to determine whether it is feasible to use 12% CRMB with NS particles while utilising seven different NS contents (NS contents of 1, 2, 3, 4, 4.5, 5, and 6%). To learn more about the properties of CR-NS binders, several experiments were carried out, including the standard tests, brookfield viscosity (BV), dynamic shear rheometer (DSR), storage stability, short term ageing, and long term ageing.

2 Goals of the Study (1) To determine Optimum Mixing Time for preparation of Crumb rubberNanosilica modified bitumen. (2) To evaluate the Optimum Nanosilica content for the Crumb rubber modified binder (CRMB) using softening point method and rheological parameters (G* and δ). (3) To investigate the influence of Crumb rubber-Nanosilica nanocomposite on viscosity of the asphalt binder. (4) To study the effect of Crumb rubber-Nanosilica content on rheological behavior of the asphalt binder based on rutting and fatigue. (5) To evaluate the effect of Crumb rubber-Nanosilica on Ageing and high temperature storage stability of the asphalt binder.

102

T. Mehraj et al.

Table 1 Base binder physical properties [8] Test

Standard code

Values

Specification limit (minimum)

Softening point (°C)

IS: 1205

46

40

Ductility (cm)

IS: 1208

100+

75

Penetration 0.1 mm at 25 °C

IS: 1203

88

80

Dynamic viscosity at 60 °C

1S: 1206(Part II)

1064

800

Kinematic viscosity at 135 °C

1S: 1206(Part III)

278

250

Table 2 Crumb rubber physical properties

Specification

Value

Particle size

0.72

>7

< 16.88

Unsafe

< 0.72

16.88

Safe

> 1.04

>9

< 19.29

Unsafe

< 1.04

19.29

Safe

> 2.24

> 2.55

< 30

Unsafe

< 2.24

< 2.55

> 30

Safe

> 2.92

> 3.2

< 32.79

Unsafe

< 2.92

< 3.2

> 32.79

Safe

> 1.2

> 9.35

< 18

Unsafe

< 1.2

< 9.35

> 18

Safe

> 1.48

> 10

< 22.5

Unsafe

< 1.48

< 10

> 22.5

Table 5 Severity level for interaction of pedestrians with all categories of vehicles Vehicle Category

Type of Interaction

All Vehicles

VPF PPF

Severity Level

Threshold by variables PET (sec)

Safe Distance (m)

Vehicle approaching speed (kmph)

Safe

> 2.28

>2

< 28.69

Unsafe

< 2.28

28.69

Safe

> 1.2

> 9.35

< 22.5

Unsafe

< 1.2

< 9.35

> 22.5

484

D. Singh et al.

6 Conclusion This study evaluated the performance of pedestrian safety using the interaction of a pedestrian-vehicle method based on video graphics data. Overall, this research provides different safety surrogate measures to define the severity of the conflict between pedestrians and vehicles in a crosswalk at uncontrolled intersections. Descriptive statistics of estimated conflict indicators provide a general idea of the risk associated with pedestrian crossings. Although cluster analysis was used to categorize these indicators according to severity, the actual risk assessment. The k-means clustering approach is used to define severity levels into four classes. The study identified a high risk of collision in VPF and PPF below PET thresholds 2.28 s and 1.2 s, respectively. The same on another SSM indicator Safe Distance (SD) found a high risk of collision below thresholds 2 m and 9.35 m. Safety of pedestrian during pedestrian vehicle interaction is highly influenced by the speed of approaching vehicles. Collision risk increases while increasing the speed of vehicles. Due to the higher speed of the vehicle, a pedestrian might not have enough time to react and avoid a collision. However, higher thresholds of PET and SD are not considered to be safe interactions. The severity of conflict is jointly attributed to the threshold value of PET & SD and the speed of the approaching vehicle. For instance, if values of PET and SD are higher than the risk of collision threshold, pedestrian-vehicle interaction might not be considered safe if the approaching vehicle speed is higher. The novelty of research is consideration of spatial and temporal SSM indication additionally with a demographic parameter such as gender of pedestrian. The future scope of this study may include analysis of pedestrian age, education affects assessments of pedestrian safety, and extension of this methodology in different traffic facilities such as four-leg intersections, mid-blocks, etc.

References 1. Marisamynathan S, Vedagiri P (2020) Pedestrian safety evaluation of signalized intersections using surrogate safety measures. Transport 35(1):48–56. https://doi.org/10.3846/transp ort.2020.12157 2. Alsop J, Langley J (2001) Under-reporting of motor vehicle traffic crash victims in New Zealand. Accid Anal Prev 33:353–359 3. Tarko AP, Davis G, Saunier N, Sayed T, Washington S (2009) Surrogate measures of safety. In: White Paper ANB20 (3) Subcommittee on Surrogate Measures of Safety ANB20 Committee on Safety Data Evaluation and Analysis Contributors. Transportation Research Board 4. Shah H, Vedagiri P (2017) Evaluation of Pedestrian Safety at Unsignalised Intersection under Mix Traffic Condition Using Surrogate Safety Measures. 15th International Conference on Computers in Urban Planning and Urban Management (CUPUM 2017) 5. Golakiya HD, Chauhan R, Dhamaniya A (2020) Evaluating safe distance for pedestrians on urban midblock sections using trajectory plots. European Transport - Trasporti Europei 2015(75):1–17

Assessment of Pedestrian Safety at Urban Uncontrolled Intersections …

485

6. Chaudhari A, Gore N, Arkatkar S, Joshi G, Pulugurtha S (2021) Exploring pedestrian surrogate safety measures by road geometry at midblock crosswalks: A perspective under mixed traffic conditions. IATSS Research 45(1):87–101. https://doi.org/10.1016/j.iatssr.2020.06.001 7. Kathuria A, Vedagiri P (2020) Evaluating pedestrian vehicle interaction dynamics at unsignalized intersections: A proactive approach for safety analysis. Accid Anal Prev 134. https:// doi.org/10.1016/j.aap.2019.105316 8. Khan T, Mohapatra SS (2021) Effect of operational attributes on lateral merging position characteristics at mid-block median opening. Transportation Letters 13(2):83–96. https://doi. org/10.1080/19427867.2019.1710037 9. Kumar A, Paul M, Ghosh I (2019) Analysis of Pedestrian Conflict with Right-Turning Vehicles at Signalized Intersections in India. Journal of Transportation Engineering, Part A: Systems 145(6):1–12. https://doi.org/10.1061/jtepbs.0000239 10. Kadali BR, Vedagiri P (2016) Proactive pedestrian safety evaluation at unprotected mid-block crosswalk locations under mixed traffic conditions. Saf Sci 89:94–105. https://doi.org/10.1016/ j.ssci.2016.05.014 11. Khan T, Mohapatra SS (2020) Influence of driver and vehicle attributes on operational characteristics of U-turning vehicles

A Review of Safety and Operational Impacts of Various Speed Limits Abhinav Mishra, A. Mohan Rao, and Darshana Othayoth

Abstract Speed is the uttermost element influencing the incidence and ferocity of road accidents. The mitigation of the frequency of over speeding is seen as an essential goal for mitigating the number and severity of collisions, and the conventional method for doing so is through posted speed zoning or posted speed limit. Various speed control strategies are now being incorporated on roads to curb accidents’ frequency and risk. Some of the speed control strategies used worldwide are Uniform Speed Limit (USL), Differential Speed Limit (DSL), Variable Speed Limit (VSL), and Lane-Based Speed Limit (LBSL). In this paper, previous research based on the implementation of DSL on different classes of roads at various road stretches has been summarised, and all the conclusions are considered for an upcoming virgin project titled “Determining safety aspects of differential speed limit on Indian roads”. Keywords Differential speed limit · Uniform speed limit · Microsimulation · Traffic safety

A. Mishra (B) · D. Othayoth Department of Civil Engineering, National Institute of Technology, Tiruchirappalli, India e-mail: [email protected] D. Othayoth e-mail: [email protected] A. M. Rao Senior Principal Scientist, Head-TES Division, CSIR—Central Road Research Institute, Mathura Road, New Delhi, PO, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_39

487

488

A. Mishra et al.

1 Introduction Speed limit with a higher compliance rate is the backbone of a safe traffic system. Accidents on the road are unwelcome events that often result in injury or death. According to the Ministry of Road Transport & Highways (MORTH) report, Road Accidents in India [21], the main reason which holds the reason behind most accidents and death on the Indian roads on every class of National Highways is over-speeding. The Nilsson Power Model states that for every 5% rise in average vehicle speed, there are 10% more accidents resulting in injuries and 20% more accidents resulting in fatalities. About 10% fewer accidents result in injuries, and 20% fewer accidents result in deaths for every 5% drop in average vehicle speed [4]. According to Haworth et al. [14], if the speed of the mobile vehicle is above 30 km/h, then the human tolerance for harm from an automobile will be surpassed. According to Mohan et al. [22], road accident is the result of many parameters. Various methodology and techniques are now been implemented to mitigate the frequency of over-speeding. There are lot of studies been carried out to compare various speed control management techniques and their impact on road safety. Those studies include some limitations which can cause irregular relationship between speed and hazard or skewness of any specific result due to inclusion or omission of any specific attribute which or may not affect the road safety explicitly or implicitly. This extensive literature reviewing is carried out to enlist the advantages and disadvantages of various methodologies which enable the future research to improve their methodologies to obtain an non skewed or non-biased conclusions.

2 Speed Limit Due to the impact of speed on roadway safety, speed limits are one of the traffic concerns in developing countries. The term ‘speed limit’ refers to the maximum speed that can be legally enforced in the interest of safety and mobility on the road, considering environmental, roadway, and roadside factors. Generally, speed limit is of three types i.e., single-speed limit, differential speed limit, and advisory speed limit. The Uniform/Single speed limit (USL) strategy instructs to implementation of an equal maximum speed limit for all classes of vehicles. In contrast, the Differential Speed Limit (DSL) strategy suggests two different maximum speed limits for two different vehicle classes, i.e., passenger class vehicles and goods carrying vehicles. The advisory speed limit is suggested speeds for bends, junctions, or other various areas where operating speed on that stretch is regulated to a lower value than maximum legal speed or implemented speed limit by characteristics of the roadway geometry. The primary parameter in establishing maximum or posted speed limits, which are often applied to all classes of vehicles, are either highway design or geometric constraints.

A Review of Safety and Operational Impacts of Various Speed Limits

489

According to Methods and Practices for Setting Speed limits: An Informational Report, GHSA [12], there are four mainstream techniques for establishing speed limits i.e., engineering approach, expert system approach, optimization, injury minimization, or safe system approach. Factors used in the determination of optimal speed limits are travel time, vehicle operating costs, road collisions, noise pollution due to the movement of traffic, and atmospheric pollution. The maximum speed limit is established in accordance with the collision types that are likely to occur in a safe system approach, the bump or thrust forces that will result from those crashes, and the human body’s capacity to withstand those forces. Apart from speed restrictions, there are other facts regarding speed to consider, such as how engineers choose a design speed to guide their design decisions and make plans before creating a new route but when road work is completed, engineers will quantify the prevailing speed by measuring the operating speed.

3 Differential Speed Limit Controversial findings can be seen in the literature regarding the use of USL or DSL. Regarding the history of implementation of DSL on various kinds of roads, many cities have different speed limits according to the type of vehicle e.g., in Missouri, cars were allowed to travel up to 70 mph, whereas trucks were limited to 60 mph. On rural national routes in Ireland, 15 mph was the differential speed limit posted in 1979 i.e., 55mph for cars and 40 mph for trucks [15]. Many studies and research have been carried out to quantify the effectiveness of DSL in reducing crashes and improving speed variance. Solomon [28] discovered a U-shaped relationship between the amount of speed deviation from the average and the frequency of collisions. Any increment in variance can prompt rise in the number of accidents, particularly accidents including (non-compliant) passenger car and (compliant) heavy vehicles. Increasing reciprocity between cars as a result of increased speed variance might result in an increase in collisions. Duncan et al. [6] discovered that a large speed difference enhances the severity of rear-end collisions between passenger vehicles and heavy vehicles.

4 Previous Studies on Speed Limit and Impact of DSL Malyshkina and Mannering [20] explored the connection between posted speed limits and the observed severity level of injuries involved in the accident to study the impact of speed restrictions on traffic safety. Their research showed that Interstate speed restrictions had no appreciable effect on the severity of accidents or injuries. Regarding the effect of compliance rate of speed limit on traffic flow and accidents, Bains et al. [3] conducted a thorough analysis of the impact of speed restriction compliance on the capacity of Indian Expressways. According to the findings,

490

A. Mishra et al.

roadway capacity increases when the percentage of cars adhering to the legal speed limit rises. In addition, travel durations are reduced somewhat as driver compliance levels rise. Gayah et al. [10] conducted research to observe the impact of posting speed limits below the advisory on mobility and safety. Heavy enforcement inside the lower speed limit zone resulted in a reduction in average speed as well as on 85th percentile speeds approximately by 4 mph and also causes increments in speed limit compliance rate. Speed limits 10 mph lower than engineering guidelines were associated with a statistically significant reduction in the incidence of both total and front-end collisions, but a statistically significant increase in the probability of events with fatal injury. Sugiyanto et al. [29] conducted study to observe the impact of lowering the speed limit on mobility and the environment due to a reduction in the speed limit. Study findings suggested that emissions of a variety of pollutants were found to be reduced at lower speeds and also hydrocarbon emissions were reduced at lower speeds, whilst Carbon Monoxide (CO) and Particulate Matter (PM) had the minimum level of exhalation at medium speeds. Yao et al. [31] conducted a study to examine the speed limit credibility and compliance on UK roads. The study concluded that the most frequent value of tendered safe speed limit was adopted as the most viable speed limit. Additionally, drivers perceive and want to travel at faster speed higher than the posted speed limit irrespective of whatever the speed limit was established. Kwayu et al. [18] used a multilevel negative binomial regression model to analyse the impact of changing the speed limit on Michigan freeways and found that increasing the speed limit by 15 mph resulted in a 21% increase in overall accidents and 11.9% increment in crashes related to injuries and fatalities which contradict the conclusions from the study carried out by Garber et al. [9] and the Department of Transportation Federal Highway Administration. They carried out a study to examine the safety aspects of the implementation of DSL strategy on rural interstate roads in the US and concluded that irrespective of which speed control strategy was implemented among USL & DSL there will be a growth in crash rate and average speed of the vehicles. Along with the study by Malyshkina and Mannering [20], Ossiander and Cummings [23] also advocated the same findings where they studied impact of freeway speed limit and traffic fatalities in Washington state using crash data and vehicle data on Washington state freeway from 1974 to 1994. They concluded that there was growth in the value of speed variance as well as in the 85th percentile speed on both rural as well as urban roadways from 1980 onwards which remain independent of any changes in speed Huang et al. [16] carried out a study to observe the effect of variation in the speed limit in the maintenance work area of Expressway during morning and night-time. Effects on the capacity of the maintenance work area come out to be different for a different proportion of big vehicles in traffic composition because more oversized vehicles showcase poor performance in terms of acceleration, deceleration, and braking compared to smaller vehicles.

A Review of Safety and Operational Impacts of Various Speed Limits

491

Garber and Gadiraju [8] focused on the impacts of DSL on the traffic speed and rate of accidents. Results concluded that the DSL strategy with not significantly more efficient than the USL strategy in mitigating the risk of exceeding the speed limit. Both strategies showcase the similar results and concluded that traffic phenomenon will fluctuate irrespective of implemented speed control strategy. Korkut et al. [17] conducted a thorough investigation of the impact of lane restrictions for trucks on traffic and crashes, as well as the implementation of the DSL strategy. Crash rates exhibit a strong relationship with variation in mean truck speed, the difference between mean speed of trucks and mean speed of passenger vehicle class, and lane occupancy. Ghods et al. [11] carried out research to observe the influence of DSL on the performance of safety aspects of two-lane highways using microsimulation methodology. The study concluded that in DSL and MSL speed control techniques, the frequency of car-to-car overtakes decreases while the number of car-truck overtakes increases. Habtemichael and Santos [13] carried out a study to provide quantitative evaluations of safety impacts and compliance of VSL on the 12-km study area of Motorway A-5 in Portugal, by simulation methodology. The study concluded that this method provided the safest advantages in heavily dense traffic, then in mildly dense traffic, and at minimum, in an uncrowded traffic state. By analysing vehicle speed characteristics in Indiana, Michigan, and Ohio, Russo et al. [26] conducted a comparative analysis incorporating USL and DSL speed control techniques. It was shown that going from 55 to 60 mph caused the mean speed and 85th percentile speed to increase by 3 to 4 mph. Using the VISSIM traffic simulation software package, Adresi et al. [1] conducted comprehensive research on the safety and traffic performance implications of the LBSL and vehicle-based (containing USL and DSL) strategies for highways. According to the results, the LBSL exhibited a larger speed variance and lane shifting capability than the other strategies. Sadat and Celikoglu [27] undertook research to examine the effects of VSL on the Istanbul motorway, comparing scenarios with the implementation of VSL strategy and in its absence. The study revealed that even at modest compliance levels, the suggested VSL systems contributed in reducing congestion levels. The congestion may be reduced by enforcing the statutory speed restrictions, since higher compliance levels generate better outcomes. Fuel efficiency is improved, and the suggested VSL system shows promise in its potential to be environmentally benign by reducing emission of CO and NOx as well as the fuel consumptions. Soriguera et al. [7] conducted a study in Barcelona to observe how low-speed limits affect traffic flow on highways. This study concluded that low-speed limits could increase lanechanging behaviour, posing a safety risk on freeways but the limitation of this study might have tilted the results towards this conclusion as all these conclusions are based on high-compliance controlled portions. Lower compliance rates would presumably lower VSL impacts hence this set of conditions needs to be looked at. Liu and Shi [19] conducted a comprehensive study to evaluate the influence of Differentiated per-Lane Speed Limit (DPLSL) on safety aspects of freeway traffic i.e., Frequency of lane changes, variation in traffic speed and hazard rate. They concluded that when traffic congestion is on higher side and space headway is minimal, drivers are more

492

A. Mishra et al.

inclined to lane changing phenomenon and Lane changing phenomenon found to be higher in USL setting in comparison to VSL. Qu et al. [24] carried out a thorough analysis of models of traffic flow and driving behaviour under VSL control, that simulated both individual driving behaviour and associated traffic flow features. Big traffic data was garbled, processed, and investigated to compare traffic data collected before and after the introduction of the VSL technique on driving behaviour analysis. The study concluded that VSL did modify the driver’s desired speed. Under VSL management, the proportion of minor headways (less than 1 s) was considerably reduced, implying that some drivers became more cautious showcasing improvement in driving condition and road safety. Costanzo et al. [5] conducted a study to determine the overall portability and conditional advantages of an instinctive VSL strategy for a highway stretch in the city of Naples. The outcomes of the study indicated a small improvement in the average speed of up to + 2.24%, suggesting a reduction in travel time on the considered road network. He also concluded that VSL can provide environmental benefits too as for all 100 carried out simulations, there was decrement of 9.54% reduction in average fuel consumption. In comparison to the non-VSL-based scenario, the VSL-based scenario observed a decrease in both the average delay as well as in the average number of stops, indicating an improvement in the travelling condition on the roadway.

5 Identification of Critical Gaps in the Literature The general limitations of every research work related to all those studies mentioned in previous sections were the inability to capture real-time data and simulate data in the original condition in which data was collected. Alonso et al. [2] conducted research to observe practices, purpose, and analysis of the efficacy of sanctions for over-speeding. The issue with this study was the inability to capture the real intention behind complying with speeding laws, as the driver was aware that he was breaking the law. Yao et al. [31] carried out a questionnaire technique to conduct a study to examine the acceptability and dependability of speed limit and compliance on UK roads. The response rate in this research was just 10%, and the questionnaire sample size was only 100. The fact that respondents who were not in a hurry took part in this survey might be a drawback. Other limitations were that multiple subcategories of parameters and groups were not considered in the model creation or in the research methodology, and any correlation between multiple variables like age, sex, driving behaviour, and driving experience weren’t considered while establishing speed limit. Also, the metering capabilities of extremely low-speed limit were not investigated in this study. Along with that, analysis was solely focused on static traffic conditions instead of dynamic traffic conditions, but static traffic does not capture full spatial– temporal traffic characteristics. The distribution of headways, particularly the fraction of smaller headways, is essential to the traffic flow’s stability. The effect of the VSL on the percentage of cars with a headway of not more than one second was studied by Qu et al. [24]. But focusing solely on the proportion of headways of less than one second

A Review of Safety and Operational Impacts of Various Speed Limits

493

will not give a full picture of the safety state of the traffic. Inclusion of other smaller headways scenarios e.g., headways of less than 1.5 s or headways of less than 2 s can be crucial when approaching the flow’s maximum capacity. In the study of Korkut et al. [17], DSL had an ambiguous effect on traffic safety since it caused collision rates to rise when trucks went above the truck speed limit but to fall when passenger cars went over the car speed limit. An irregular negative correlation between vehicle speed limit infractions and accident rates was surprising. Since the RTMS devices provided speed for all vehicles combined for each 30 s time interval, the truck and other vehicle’s speeds were estimated under diversified-vehicle circumstances. Extending the limitation of the study, previous research was not able to draw any conclusion on the relationship between the proportion of trucks in total traffic composition and the rate of accident occurrence. In a study by Garber et al. [9] and the Department of Transportation Federal Highway Administration, they were unable to distinguish how the USL/DSL strategies had an impact. The study’s analysis of aggregated speed data did not reveal any DSL-related effects. Within the parameters of the investigation, no consistent safety differences between DSL and USL were found. In the finding of Sugiyanto et al. [29], they did not able to find one single speed limit which can minimize CO, NOx , HC, PM, because for speed limit of 40 kmph the emanation of CO and NOx were observed to be at its lowest, but for speed limit of 50 kmph particulate matters were observed to be minimal. Russo et al. [26] in addition to looking into the effects of USL versus DSL on trip speed, also looked into the travel speeds of trucks, buses, and passenger cars in jurisdictions with various speed limits. However, according to the research, speed selection could have been influenced by local variables, and those local factors might have been the reason behind the variations in driver speed choices. So, the exact relationship between speed control strategy and crash rate could not be predicted. Liu and Shi [19] conducted research using DPLSL. The results of this study were heavily reliant on simulations, which was one of its limitations. For gathering the speed data and assessing the two hazardous situations, the precise locations of all cars at a given moment were required, and in simulation, traffic movement is predefined and not so erratic in comparison to real traffic conditions. Real traffic flow, on the other hand, is irregular and haphazardous, and hence this was termed as a limitation of this study. Most of previous research on the impact of DSL on safety aspects of traffic and roads used statistical methodologies like the before-and-after approach, which might not reflect the true improvement or might not show the reasoning behind any changes, which can be included as limitations of previous studies.

6 Conclusions One of the main inspirations for this literature review was to explore the efficacy of DSL on speed and safety aspects of operating conditions on the road. The majority of the previous studies related to this pre-identified strategy were unable to pinpoint which speed control strategy among USL and DSL gives better performance in

494

A. Mishra et al.

terms of their impact on road safety. Many drivers will disregard the posted speed limit if it is too low, resulting in common following and acceptance rates. Slower speeds reduce the intensity of collisions; however, higher speed variance inflated the likelihood of collisions. Most of the studies concluded that there will be growth observed in average speed of vehicles and accident rate with time irrespective of whatever speed control strategies have been implemented. A study on the impact of low-speed limits on traffic flows on highways concluded that low-speed limits could increase lane-changing behaviour, posing a safety risk on freeways. Also, other studies on low-speed limits suggested that emissions of various pollutants were found to be reduced at lower speeds and hydrocarbon’s emissions were reduced with slower speed, whilst CO and PM have the minimum ejection rate at moderate or at the usual speeds. Stating these findings in more detailed perspective, previous study did not able to find one single speed limit which can mitigate CO, NOx , HC, PM, as they found out that CO and NOx were at the lowest emission level. The effects of a 5 kmph reduction in travel speed are greater at lowers speeds, and a reduction of speed limit to 50 kmph brings a substantial decrement in the risk of crashes or collisions and fatalities. Furthermore, few studies concluded that the VSL strategy could cause a reduction in the proportion of smaller headways (less than 1 s), which points towards the positive impact of VSL with respect to safety impact on traffic, and implicitly, it can be said that drivers became more cautious under VSL strategy. So, summing up the previous study related to DSL and its impact different to of USLs, previous studies didn’t able to segregate out any major differences between impact of USL and DSL on safety aspects hence more dwelling in the matter needs to be undertaken. This thorough review of methodologies and pre-established practises involved in managing the speed and enhancement of compliance rate with respect to speed limit will pave a way for understanding the factors and components which implicitly or explicitly does affect the safety of road environment and its users. Further research needs to be undertaken, incorporating other reliable, realistic factors, and instead of incorporating statistical techniques for formulating a relationship between crash rate and different speed control strategies like USL and DSL One of the most significant flaws in the statistical technique is the constraints imposed on the analysis due to the lack of data hence to counter that a microsimulation study can be proposed which can replicate the real field condition. This extensive literature review will pave the way and can build a solid platform for an upcoming proposed study with a pre-identified methodology (see Fig. 1) and titled as “Determining Safety Aspects of Differential Speed Limit (DSL) on Indian Roads” in CSIR-Central Road Research Institute (CRRI), New Delhi, India.

A Review of Safety and Operational Impacts of Various Speed Limits

495

Literature Review Data Collection

Road Inventory

Traffic Characteristics

Selection of Study Stretch Road Crashes

Preparation of Simulation Data

Road Network

Vehicle Fleet

Traffic Control

No Calibration & Validation

Running Simulation and Data Extraction

Formulation of Speed Control Measures

Identification of KPI Evaluation of Speed Control Measures Suggestion on Best Speed Control Measures Fig. 1 Methodology adopted for the study titled as "Determining Safety Aspects of DSL on Indian Roads

References 1. Adresi M, Baghalishahi AM, Zeini M, Khishdari A (2016) Impact of speed limit method on motorway safety. Gradevinar 68(9):705–713 2. Alonso Plá FM, Esteban Martínez C, Calatayud Miñana C, Sanmartín J (2013) Speed and road accidents: behaviors, motives, and assessment of the effectiveness of penalties for speeding. Am J Appl Psychol, 1(3): 58–64 3. Bains MS, Bhardwaj A, Arkatkar S, Velmurugan S (2013) Effect of speed limit compliance on roadway capacity of Indian expressways. Procedia Soc Behav Sci 104:458–467 4. Cameron MH, Elvik R (2010) Nilsson’s Power Model connecting speed and road trauma: Applicability by road type and alternative models for urban roads. Accid Anal Prev 42(6):1908– 1915 5. Di Costanzo L, Coppola A, Pariota L, Petrillo A, Santini S, Bifulco GN (2020) Variable Speed Limits System: A Simulation-Based Case Study in the city of Naples. In 2020 IEEE International Conference on Environment and Electrical Engineering and 2020 IEEE Industrial and Commercial Power Systems Europe (EEEIC/I&CPS Europe) (pp. 1–6). IEEE (2020) 6. Duncan CS, Khattak AJ, Council FM (1998) Applying the ordered probit model to injury severity in truck-passenger car rear-end collisions. Transp Res Rec 1635(1):63–71

496

A. Mishra et al.

7. Francesc Soriguera F, Martínez I, Sala M, Menéndez M (2017) Effects of low-speed limits on freeway traffic Flow. Transportation Research Part C: Emerging Technologies 77:257–274 8. Garber NJ, Gadiraju R (1992) Impact of differential speed limits on the speed of traffic and the rate of accidents. Transp Res Rec 1375:44–52 9. Garber N, Miller JS, Yuan B, Sun X (2005) The Safety Impacts of Differential Speed Limits on Rural Interstate Highways. Virginia Transportation Research Council and US Department of Transportation, Federal Highway Administration, Publication No. FHWAHRT-05–042, Washington DC (2005). 10. Gayah VV, Donnell ET, Yu Z, Li L (2018) Safety and operational impacts of setting speed limits below engineering recommendations. Accid Anal Prev 121:43–52 11. Ghods AH, Saccomanno F, Guido G (2012) Effect of car/truck differential speed limits on two-lane highways safety operation using microscopic simulation. Procedia Soc Behav Sci 53:833–840 12. Governors Highway Safety Association: Speed Limit Laws (2013). 13. Habtemichael FG, de Picado Santos L (2013) Safety and operational benefits of variable speed limits under different traffic conditions and driver compliance levels. Transp Res Rec 2386(1):7–15 14. Haworth N, Tingvall C, Kowadlo N (2000) Review of best practice road safety initiatives in the corporate and/or business environment. Monash University Accident Research Centre Reports 166:119 15. Hearne R (1981) Car and Truck Speeds Related to Road Traffic Accidents on the Irish National Road System. Proc., International Symposium on the Effects of Speed Limits on Traffic Accidents and Fuel Consumption, Dublin, Ireland (1981) 16. Huang Y, Fang D, Tan F, Tao M, Si D, Shu Y (2020) Study on the Speed Limit at Night in Expressway Maintenance Area. In IOP Conference Series: Earth and Environmental Science (Vol. 619, No. 1, p. 012097). IOP Publishing (2020) 17. Korkut M, Ishak S, Wolshon B (2010) Freeway truck lane restriction and differential speed limits: crash analysis and traffic characteristics. Transp Res Rec 2194(1):11–20 18. Kwayu KM, Kwigizile V, Oh JS (2020) Assessing the safety impacts of raising the speed limit on Michigan freeways using the multilevel mixed-effects negative binomial model. Traffic Inj Prev 21(6):401–406 19. Liu M, Shi J (2018) Exploring the impact of differentiated per-lane speed limits on traffic safety of freeways with considering the compliance rate. J Adv Transp 20. Malyshkina NV, Mannering F (2008) Effect of increases in speed limits on severities of injuries in accidents. Transp Res Rec 2083(1):122–127 21. Ministry of Road Transport and Highways. Road accidents in India (2019). 22. Mohan D, Khayesi M, Tiwari G, Nafukho FM (2006) Road traffic injury prevention training manual. World Health Organization 23. Ossiander EM, Cummings P (2002) Freeway speed limits and traffic fatalities in Washington State. Accid Anal Prev 34(1):13–18 24. Qu X, Li L, Yi Z, Mao P, Yang M (2020) Traffic flow modeling of freeway variable speed limit control based on the big data of driving behavior. J Adv Transp 25. Qu Z (2009) Cooperative control of dynamical systems: applications to autonomous vehicles. Springer Science & Business Media 26. Russo BJ, Rista E, Savolainen PT, Gates TJ, Frazier S (2015) Vehicle speed characteristics in states with uniform and differential speed limit policies: comparative analysis. Transp Res Rec 2492(1):1–9 27. Sadat M, Celikoglu HB (2017) Simulation-based variable speed limit systems modelling: an overview and a case study on Istanbul freeways. Transportation research procedia 22:607–614 28. Solomon DH (1964) Accidents on main rural highways related to speed, driver, and vehicle. US Department of Transportation, Federal Highway Administration 29. Sugiyanto G, Jajang SMY (2019) The impact of lowering speed limit on mobility and the environment. In AIP Conference Proceedings (Vol. 2094, No. 1, p. 020019). AIP Publishing LLC (2019).

A Review of Safety and Operational Impacts of Various Speed Limits

497

30. Wolshon B, Pande A (2016) Traffic engineering handbook. John Wiley & Sons 31. Yao Y, Carsten O, Hibberd D (2020) A close examination of speed limit credibility and compliance on UK roads. IATSS research 44(1):17–29

Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic P. Vijayalakshmi, Nitin Kumar, and Vivek R. Das

Abstract In the Indian scenario, mixed traffic condition is seen, all vehicles use one lane, and the walking of pedestrians is a common sight being observed at the intersection. At an unsignalized junction, vehicles usually ignore lane discipline and rules of priority and tend to cross the junction without considering existing traffic. Due to this action, there is a risk of an accident, which affects the vehicle movement and capacity of the intersection. Therefore, it is very important to study driver behaviour which is very complex and it is influenced by traffic and vehicle characteristics. This paper deals with driver behaviour at unsignalized intersections analysed using fuzzy logic. Data from the study area is collected from videos recorded at an unsignalized T intersection for a duration of 1 h during peak hour. Recorded videos are played again for data extraction. Data like vehicle count and type, approach speed, size of gaps, accepted and rejected gaps are known. Models are developed in fuzzy logic using MATLAB with input as accepted gaps, vehicle speed and vehicle type to get the output of drivers’ choice as accepting or rejecting the gap. 50% of the data is used for modeling and model evaluation is carried out using 50% of the data. The models developed can be applied to different intersections with similar characters. Keywords Unsignalized intersection · Critical gap · Fuzzy logic · Gap acceptance

P. Vijayalakshmi (B) Department of Construction and Highway Technology Dayananada Sagar College of Engineering, Kumar Swamy Layout Bangalore Karnataka, Bengaluru 560078, India e-mail: [email protected] N. Kumar Department of Civil Engineering, Dayananada Sagar College of Engineering, Kumar Swamy Layout Bangalore Karnataka, Bengaluru 560078, India e-mail: [email protected] V. R. Das Department of Civil Engineering, MS Ramaiah Institute of Technology, MSR Nagar Bangalore Karnataka, Bengaluru 560054, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_40

499

500

P. Vijayalakshmi et al.

1 Introduction An intersection is a place where traffic of different directions meets, converge and move in the direction of their destination. The junction where vehicles and pedestrian movement is controlled by signals or by traffic police is “Signalized Intersection”. The intersection where vehicle and pedestrian movement is not controlled neither by signals or by traffic is a police “Unsignalized Intersection”. Since there is no objection at unsignalized intersections the drivers or pedestrians tend to cross the junction as soon as possible without considering the existing traffic. Due to improper movements, there is a risk of accidents and injuries, by which unsignalized intersections are proven to be dangerous hotspots. Since driver behaviour is complex and mixed with uncertainties less study is carried out unsignalized intersections. Other than HCM 2010 we don’t have an appropriate or standard procedure to evaluate unsignalized intersections in India. So, to solve this problem driver behaviour at unsignalized intersection need to be studied. Fuzzy theory can be used under situations involving uncertainties. nowadays fuzzy theory is used in various applications like traffic signal control, transport planning, braking system, air conditioning, qualitative and quantitative analysis, etc. Driver behaviour is complex and cannot be concluded. Fuzzy theory is a good tool to assess the behaviour of driver at unsignalized intersections. Capacity, delay, level of service or any other characteristics of intersection can be obtained by accepted gaps. therefore, Gap acceptance is used to study the driver behaviour at unsignalized intersections. Gap is the smallest time gap available to cross to the junction safely between two vehicles. If a particular vehicle accepts the gap and crosses the junction, then it’s “Gap accepted”. If the vehicle rejects the gaps and waits for a sufficient time interval to cross the junction is known as “Gap rejectance”. The acceptance of the gap depends on various factors like size of gap, type of vehicle, speed of vehicle, driver age, driver gender geometry of intersection and pedestrian movement. The final decision of accepting gaps depends on these factors which the human brain has to perceive, analyse and respond to which is very complex. Fuzzy logic theory can be used to make a decision of accepting or rejecting the gaps by considering various factors like size of gap, type of vehicle and speed of the vehicle. Models of gap acceptance are developed using Fuzzy logic with the best fit of available data.

2 Literature Review Gap acceptance is an important objective characteristic of the behavior of drivers and it has an influence on capacities, delays, and road safety at unsignalized intersections. Then adaptive neuro-fuzzy logic was applied with three input variables size of gap, subject vehicle type, and conflict vehicle type and output variable driver’s decision indicating acceptance or rejection of the gap is obtained for right turning vehicles in a study by Sangole and Patil (2014). In a study by Manish Dutta and Mokaddes Ali

Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic

501

Ahmed for TW and Four wheelers, statistical analysis and equations were developed. It has given critical gap values of aggressive and non-aggressive driver behaviour. Erratic maneuvers in the intersection area and aggressive driving are two common behaviors of drivers observed at these intersections (2017). The age of a driver plays a very important role in the gap acceptance process as concluded in a study by Harsh and Akhilesh (2017) and Sangolea and Patil (2011). Most of the studies carried out unsignalized intersections concluded that the age of drivers is important factor in determining gap acceptance behaviour. Keeping all the above parameters, we decided to consider the speed of the vehicle and gaps accepted by vehicles to analyse the gap acceptance behaviour at unsignalized uncontrolled junctions.

3 Data Collection For the collection of data T arm in Rajarajeswari Nagar with heterogeneous traffic is selected in Bangalore, the Capital city of Karnataka, one of the southern states of India for study and video graphic survey is carried out at these locations for a duration of 1 h during peak hour. The intersection is selected on the basis that it lies on plain topography, no bus bay and no parking at the intersection and vehicle movement is continuous. And suitable place is also selected near the junction to place a camera covering all movements clearly to carry out a videographic survey. The selected intersections of major and minor roads are of 4-lane two-way divided carriageway 14 m with a median of 0.5 m. A videographic survey is carried out from 10:00 AM to 11:AM for 60 min on weekdays. Recorded video is played several times to extract data like vehicle movement, vehicle count, intersection crossing time, gap size, gaps accepted and gaps rejected. Two-wheelers, car, auto, LCV, HCV and bus are majorly observed types of vehicles in the intersection. Two-wheeler composition is highly observed and cycles and buses are least observed. 93 Pedestrians are observed crossing the intersection with a speed of 1.06 m/s (Fig. 1). Fig. 1 Geometry of study area

502

P. Vijayalakshmi et al.

Fig. 2 T intersection in Rajarajeshwarinagar, Bangalore

4 Data Analysis Data consist of 4241 vehicles per hour. The total composition of traffic is composed of 62% of TW, 26% of cars, 8% of autos, 2 of LCVs, 1% of HCVs, 0.45% of bus and less than 1% of cycle type of vehicles movements observed in the recorded video. The average speed of vehicles is in the range of 13kmph -25kmph. The distance and time of intersection crossing are noted down and the speed of each type of vehicle is calculated. Gaps accepted and gaps rejected by each type of vehicle are also extracted. Critical Gaps are calculated using the Raff method. Among the various methods available for the calculation of the Critical gap, the Raff Method is selected because of its ease of carrying out the study. The sum of the cumulative no. of gaps accepted and gaps rejected are plotted the intersection of these curves is taken as a Critical Gap. From the Raff method, the Critical Gap we got for through moving TW is 1.7 s, car is 1.7 s, Auto is 1.7 s, LCV is 2.3 s, HCV is 2.5 s, and Bus is 2.8 s. And for right turning TW is 1.7 s, car is 1.8 s, Auto is 1.6 s, LCV is 3 s, HCV is 2.5 s, and Bus is 2.8 s (Fig. 2).

5 Descriptive Analyses of Speed and Gap accepted The descriptive analysis is carried out using EASYFIT Software to describe, interpret and analyse the data. Descriptive analysis helps to know about mean, median, variance, variance, skewness, etc., which is helpful in sensing the data to make conclusions (Tables 1, 2, 3 and 4).

Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic

503

Table 1 Statistical analyses of speed of moving vehicles Statistical parameters

TW(M)

TW(F)

CAR

AUTO

LCV

HCV

BUS

Range

50

49

44

33

30

8

16

Mean

24.57

24.81

21.65

21.09

21.82

22.57

15.5

Median

24

24

21

20

23

22

15

Variance

69.37

73.22

74.30

60.99

65.49

8.95

36.5

Std. deviation

8.33

8.56

8.62

7.81

8.09

2.99

6.04

Coef. of variance

0.34

0.35

0.40

0.37

0.37

0.13

0.38

Skewness

0.3

0.54

0.31

0.40

0.05

0.10

0.34

Min.%

7

8

3

7

8

19

9

Max.%

57

57

47

40

38

27

25

Table 2 Statistical analyses of gaps accepted by moving vehicles Statistical parameters

TW(M)

TW(F)

CAR

AUTO

LCV

HCV

BUS

Range

9.16

10.54

11.99

9.28

9.04

2.11

3.02

Mean

3.94

3.79

4.26

3.68

3.95

3.35

3.026

Median

4

3.5

4.15

3.5

4

3.3

3.24

Variance

3.24

3.60

2.77

2.90

2.98

0.15

0.58

Std. deviation

1.80

1.90

1.66

1.70

1.72

0.39

0.76

Coef. of variance

0.45

0.50

0.39

0.50

0.44

0.11

0.25

Skewness

0.80

1.02

0.18

0.89

0.80

0.36

1.15

Min.%

1

1

1.01

1

1.12

2.01

1.22

Max.%

10.16

11.54

13

10.28

10.14

4.12

4.24

Table 3 Statistical analyses of speed by right turning vehicles Statistical parameters

TW(M)

TW(F)

CAR

AUTO

LCV

Range

38

27

25

25

22

Mean

15.56

17.80

12.53

15.92

13.66

Median

15

17

12

16

13

Variance

47.21

38.47

30.53

35.63

38.17

Std. Deviation

6.87

6.20

5.53

5.97

6.18

Coefficient of Variance

0.44

0.35

0.44

0.37

0.46

Skewness

0.67

0.79

0.65

0.04

0.58

Min. %

4

5

3

5

5

Max. %

42

32

28

30

27

504

P. Vijayalakshmi et al.

Table 4 Statistical analyses of gaps accepted by right turning vehicles Statistical parameters

TW(M)

TW(F)

CAR

AUTO

Range

55.85

10.54

11.99

9.83

LCV 9.76

Mean

4.21

3.79

4.33

3.63

3.96

Median

4

3.56

4.18

3.45

3.95

Variance

18.51

3.57

2.87

2.94

3.41

Std. deviation

4.30

1.89

1.69

1.72

1.84

Coefficient of variance

1.02

0.50

0.39

0.47

0.47

Skewness

9.07

1.02

0.11

0.85

0.90

Min.%

0.15

1

1.01

Max.%

5.6

11.54

13

0.45

1.12

10.28

10.88

6 Distribution Fit for a Data The data were subjected to distribution fittings in the Easy fit software for both inputs (Gap Accepted and Crossing Speed) that are planned to be used as inputs in the MATLAB fuzzy logic. Burr, General Extreme value, Gamma, Pearson, Pareto, Dagum, Cauchy. In K-S distribution analyses for Speed, the General Extreme value showed the best fit most of the time, therefore we have used the General Extreme Value and for Gaps accepted, Dagum value showed the best fit most of the time, therefore we have used Dagum distribution as a membership function in MATLAB. Table below shows the distribution parameters of speed and Gaps Accepted in Easy Fit software (Tables 5 and 6). Table 5 Distribution data analysis of Speed Vehicle type

Distribution

K-S

Parameters

Static

Rank

TW

Burr

0.0375

1

K = 6.373, α = 3.536, β = 44.77

CAR

Gen Gamma

0.0430

1

K = 2.929 α = 0.776 β = 25.61 γ = 1.518

AUTO

Pearson6

0.0578

1

α1 = 7.57 α2 = 4.369E + 7 β = 1.22E + 8

LCV

Gen.Pareto

0.0886

1

K = –0.8969 σ = 25.71 μ = 8.129

HCV

Cauchy

0.1785

1

σ = 2.419 μ = 22.40

BUS

Johnson SB

0.1506

1

γ = 0.2356 δ = 0.280 λ = 15.87 ξ = 8.891

Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic

505

Table 6 Distribution data analysis of gap accepted Vehicle type

Distribution

K-S

Parameters

Static

Rank

TW

Normal

0.056

1

σ = 1.74 μ = 3.94

CAR

Dagum

0.0522

1

K = 0.255 α = 6.96 β = 4.79 γ = 0.93

AUTO

Weibull

0.0560

1

α = 2.511 β = 4.107

LCV

Johnson SB

0.0622

1

γ = –0.201 δ = 0.939 λ = 6.040 ξ = 0.3081

HCV

Inv. Gaussian

0.074

1

λ = 24.61 μ = 3.348

BUS

Dagum

0.1041

1

K = 0.1728 α = 3.041 β = 37,735 γ = 3–37,732

7 Gap Acceptance Fuzzy Models Gap Acceptance models are created to study the behaviour of drivers. Since human behaviour is very complex and uncertain, fuzzy tools are used to analyse this type of situation. Fuzzy logic in MATLAB is used to analyse the data and generate the models. The models are developed for different types of vehicles using speed and gap accepted by particular types of vehicles. Fuzzy logic toolbox is a six-layer interface system with input, fuzzification layer, rules layer, normalization layer, defuzzification layer and output layer. In our study Input 1 is taken as gaps accepted and Input 2 is taken as the speed of vehicles to get the output of the drivers choice with NO as rejecting the gaps and YES as accepting the gaps. The distributions fits for gaps accepted and speed values of extracted data are determined using EASY FIT software. In Easy Fit software, the individual values are used and graphs are plotted to check the suitable distribution fit. Among the plotted curves General extreme value distribution shows the best fit for the speed of the vehicle and the Dagum distribution shows the best fit for gaps accepted in the K-S method. These distribution functions are applied to the Membership function in fuzzy logic with Low, Medium, and High-value ranges. Then 9 rules to process the input are applied in the rules layer. Rules are. If “Gap” is LOW, “Speed” is LOW, then Choice is “NO”. “Gap” is LOW, “Speed” is MEDIUM, then Choice is “NO…. So, on. The results were obtained in the form of surfaces and rules for TW and Car are shown below Figs. 3 and 4.

506

P. Vijayalakshmi et al.

Fig. 3 Gap acceptance model for TW

Fig. 4 Gap acceptance model for Car

8 Conclusion . The average speed of two wheelers is 25 kmph, Car is 22 kmph, Auto is 21 kmph, LCV is 22 kmph, HCV is 23 kmph and Bus is 15 kmph. . The average gaps accepted of Two-wheelers are 1.7 s, Car is 1.7 s, Auto is 1.7 s, LCV is 2.3 s, HCV is 2.5 s and for the Bus is 2.8 s. . Distribution Fit for speeds of different vehicles we got using Easy Fit Software are Burr, General Extreme value, Gamma, Pearson, Pareto, Dagum, Cauchy, Chi-Squared, Johnson SB, Pert, Chi-Square. Out of these distributions, In K-S distribution analyses for Speed, General Extreme value showed the best fit. . Distribution Fit for Gaps accepted for different vehicles we got using Easy Fit Software are General Extreme value, Gamma, Pearson, Pareto, Dagum, Cauchy, Chi-Squared, Johnson SB, Pert, Chi-Square, Gum bell, Nakagami, Weibull, and

Study on Driver Behaviour at Unsignalized Intersection Using Fuzzy Logic

507

log Pearson. Out of these distributions, In K-S distribution analyses for gaps accepted, Dagum value showed the best fit. . Gap acceptance models are developed for different types of vehicles in MATLAB to study the driver’s behaviour

References 1. Annual Accidents reports, MORTH 2020 2. Akhilesh Kumar Maurya, Harsha J Amin, Arvind Kumar “ Estimation of Critical Gap for through movement at four leg uncontrolled Intersection” 11thTransportation Planning and Implementation Methodologies for Developing countries https://doi.org/10.1016/j.trpro.2016. 11.076 ,TPMDC – Elsevier 2014 3. Ashalatha R, Chandra S (2011) Critical Gap through clearing behaviour of drivers at unsignalized Intersections. Springer 2011. 4. Gopal RP, Jayant PS (2015) Behavior of two-wheeler at limited priority uncontrolled T intersection. Springer 2015. 5. Gowri A, Anuroop C (2016) Analysis of occupation time of vehicles at urban unsignalized intersections in non-lane-based mixed traffic conditions. Springer 9th July 2016. 6. Harsh JA, Maurya AK (2015) Modelling the Gap Acceptance Behaviour of Drivers of TwoWheelers at Unsignalized Intersection in Case of Heterogeneous Traffic Using ANFIS, October 2015 7. Manish D, Ahmed MA (2017) Gap acceptance behaviour of drivers at uncontrolled Tintersections under mixed traffic conditions Springer 15th December 2017 8. Mohan M, Chandra S (2020) Analysis of Driver behaviour at unsignalized intersections. J Indian Road Congress

Capacity Analysis and Safety Assessment of Unsignalized Intersection Using Conflict Technique P. H. Souparnika, Sheela Alex , and Padmakumar Radhakrishnan

Abstract Intersections pose special safety concerns because of the high probability of critical conflicts resulting from unsafe driver actions and maneuvers. The absence of movement priorities, lack of lane discipline, forced entry by non-priority movements, etc., at unsignalized intersections violate the assumptions involved in capacity calculation. A key aspect of this study is to analyze conflicting flows at unsignalized intersections, which vary depending on site conditions and geometrical features in heterogeneous traffic environments and to estimate movement capacity at unsignalized intersections. At unsignalized intersections, vehicles from different directions cross and turn simultaneously, resulting in severe vehicle-vehicle conflicts. Thus, traffic safety is an important aspect to be evaluated at unsignalized intersections. The study also aims at conducting a safety assessment at unsignalized intersections using the microsimulation model (VISSIM) coupled with the Surrogate Safety Assessment Model (SSAM). The study found that SSAM is adaptable to intersections of varying geometry and serving heterogeneous traffic. Keywords Unsignalized intersection · Capacity · Vissim · Ssam

P. H. Souparnika (B) · S. Alex · P. Radhakrishnan Department of Civil Engineering, College of Engineering Trivandrum, Thiruvananthapuram, Kerala, India e-mail: [email protected] S. Alex e-mail: [email protected] P. Radhakrishnan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_41

509

510

P. H. Souparnika et al.

1 Introduction Unsignalized intersections frequently see serious conflicts between crossing traffic due to erratic traffic maneuverability. Drivers are highly aggressive; therefore, the rules of priority are often ignored at unsignalized intersections and also lane discipline is nearly non-existent. Hence, the capacity estimation procedure followed in developed countries is inapplicable to intersections having mixed traffic conditions. The study intended to estimate movement capacities at unsignalized intersections under heterogeneous traffic. For the capacity analysis at unsignalized intersections, conflicting flow is determined using the volume of traffic and relative levels of hindrance brought by higher priority movements to the subject movement. Indo HCM [1] explains in detail how the conflicting flow computation varies based on intersection geometry and the relative influence of conflicting movements on subject movement. With this motivation, the study examines how the conflicting flow varies based on site geometry and the influence of different conflicting movements on the subject stream at field, to compute movement capacity at unsignalized intersections. The major advantage of this traffic conflict technique (TCT) based analysis is that it examines real-time interactions of traffic prior to the occurrence of a crash. The conventional approach of crash data analysis includes many drawbacks such as long observation period, lack of crash data, and inaccurate information about the crash patterns. Surrogate safety measures (SSM) were used as a substitute for traditional crash data analysis for a more accurate and rapid evaluation of safety at the site [2]. Therefore, safety assessment is also conducted at selected locations using the surrogate safety measures. The objective of the study is to determine conflicting traffic flow rates based on site geometry and the relative impact of conflicting movements on subject vehicle and to compute site-specific capacity of unsignalized intersections using conflict technique. Also, the vehicle conflicts at selected unsignalized intersections were determined and categorized into various types using the microsimulation model, VISSIM coupled with Surrogate Safety Assessment Model (SSAM).

2 Literature Review Joewono et al. [3] conducted an analysis of unsignalized intersections with conflict method for capacity estimation under mixed traffic conditions of non-priority intersections. The bending of right-of-way at unsignalized intersections causes the different priority ranking of traffic streams and provides a modification of procedure to determine the volume of the priority stream for capacity calculation [4]. Ghanim and Shaaban [5] investigated the feasibility of using SSAM to identify and classify traffic conflicts between vehicles and pedestrians by analyzing simulated trajectories. Mohan and Chandra [6] focused on developing the entire procedure for estimating the capacities of movements at unsignalized intersections. Using Harder’s capacity model as a base, the procedure to estimate the parameters of this model

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

511

was revised to suit the traffic operations in developing countries. Arkatkar et al. [7] conducted a safety evaluation of unsignalized intersections using PET for through movement as well as turning movements and also analyzed the critical speed of the conflicting vehicles at the intersection for assessing the extremity of conflicts using PET values. Safety assessment at intersections in urban areas is conducted using a micro-simulation model, VISSIM coupled with the Surrogate Safety Assessment Model (SSAM) [8]. The number, type and severity of conflicts were determined in the study and conflicts were categorized into three types according to the conflict’s angle, as rear-end, lane change and crossing conflicts. Goyani et al. [9] attempted to investigate crossing conflicts at unsignalized T-intersections under mixed traffic conditions using post-encroachment time (PET) as a surrogate safety measure (SSM). Paul and Ghosh [10] proposed an approach to determine the PET threshold for carrying out an efficient and faster safety evaluation at unsignalized intersections under highly heterogeneous traffic conditions. From the literature reviewed, it is clear that limited studies analyzed the capacity of unsignalized intersections by conflict technique and recent studies explored the viability of using Surrogate Safety Measures (SSMs) to identify and categorize traffic conflicts between vehicles by evaluating simulated trajectories.

3 Methodology The entire procedure adopted for this study is depicted in the form of a flowchart as shown in Fig. 1.

3.1 Site Selection The intersection having low to medium flow with sufficiently flat gradients possessing three-legged as well as four-legged configurations were considered for the study. Fig. 2. shows the Vazhuthacaud junction and Sreekaryam junction at Trivandrum city and Chadayamangalam junction situated at Kollam.

3.2 Data Collection The traffic data were collected from 1.00 pm to 2.00 pm in a day during off-peak hours at each unsignalized intersection using a video-graphic survey. A geometric survey was also conducted to collect road inventory details. Table 1 shows the geometric details of selected intersections.

512

P. H. Souparnika et al.

Fig. 1 Flowchart of the methodology

(a)

(b)

(c)

Fig. 2 Study locations: a Vazhuthacaud Junction b Sreekaryam Junction c Chadayamangalam Junction Table 1 Geometric details of the selected site Intersection

Intersection type

Major street configuration

Carriageway width (m)

Vazhuthacaud

Three-legged

Four-lane divided

18

Chadayamangalam

Three-legged

Two-lane undivided

7.5

Sreekaryam

Four-legged

Two-lane undivided

7.5

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

(a)

513

(b)

Fig. 5 Vehicular movements at a Three-legged intersection b Four-legged intersection

3.3 Data Extraction and Analysis From the collected video, vehicle composition, classified volume count, PET values, and gap and lag for different vehicle classes for critical gap estimation were extracted. Vehicle Categories and Composition Figure 5. shows the vehicular movements at three-legged and four-legged intersections. Traffic volume and composition were extracted from the video-graphic recording by a manual count at three unsignalized intersections. The volume of traffic executing each turning movement at the intersection is converted into the equivalent number of passenger cars using the PCU values obtained from [1] and presented in Table 2. Determination of Post-Encroachment Time (PET) Post-Encroachment Time (PET) is the time difference between when the first vehicle leaves the conflict area, and the time when the second vehicle enters the conflict area. The data extraction was done manually from the video using the grid line technique. Kinovea™ software was used to extract PET values from the field. The PET values are calculated by noting down two time events T1 and T2. PET = T2 − T1

(1)

514

P. H. Souparnika et al.

Table 2 Traffic converted into PCUs Vazhuthacaud junction Movement

2W

CAR

3W

LCV

HCV

Total (PCU/hr)

1

169

367

126

31

5

698

2

270

426

179

49

7

931

5

240

482

160

71

5

958

6

71

162

36

17

5

291

7

110

209

87

7

5

418

9

66

251

5

14

5

421

Chadayamangalam junction Movement

2W

CAR

3W

LCV

HCV

Total (PCU/hr)

1

66

92

86

71

19

334

2

94

328

141

126

124

813

5

109

399

136

136

141

921

6

57

56

86

14

19

232

7

45

80

75

27

10

237

9

46

100

102

41

10

299

Sreekaryam junction Movement

2W

CAR

3W

LCV

HCV

Total (PCU/hr)

1

116

241

64

27

12

460

2

239

412

114

82

50

897

3

14

36

31

20

101

4

23

28

20

9

80

5

135

422

106

88

6

27

60

24

26

137

7

50

80

24

37

191

8

63

65

30

41

9

122

221

67

27

10

22

42

31

21

11

61

72

30

24

12

17

36

11

44

133

884

199 29

466 116

14

201 108

where T1 = Time, when the offending vehicle leaves the conflict area; T2 = Time, when the conflicting vehicle enters the conflict area. Fig. 4. shows the screenshot of PET extraction. The extracted PET values show which conflicting movements are critical, i.e. PET< 1.0 s at field [10] and based on that, the impact of conflicting movements on subject movement can be found out and also number of critical conflicts existing at site. Determination of Conflicting Flow Rates and Capacity Estimation

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

515

Fig. 4 Screenshot of PET extraction in Kinovea software

Conflict occurs between two movements at an intersection when they compete for a common right of way and conflicting flow refers to the cumulative volume of traffic from different streams that will have an impact on the operation of a non-priority movement at an intersection. Different weights were assigned ranging from 0 to 2 for different conflicting streams to account for the impact of these movements on the subject stream [1]. A weightage of ‘0’ implies having no significant influence, while a weightage of ‘2’ indicates an extensive effect on the driver’s gap acceptance behavior. These weights were assigned on the basis of site geometry and the relative impact of conflicting movements on subject movement at the field [1]. Any movement of higher priority with which the subject movement shares the right of way is included as a conflicting flow for that subject movement. In India, priority among movements is rarely followed, thus movements of lower priority might cause hindrance to the movements of higher priority. Hence, lower priority movement is also to be considered while calculating the rate of conflicting flow. Table 3 shows the conflicting flow computation at selected locations. From the field observation, it is clear that through traffic on the major street was observed to have the highest priority owing to their higher speed and ease of clearing the conflict area. Right turners from major and minor streets experience a high risk of conflict from through movements from the major street. Through and leftturning movements from major and minor streets enjoy the highest priority and are unimpeded by any other movements. In calculating conflicting flow at three-legged intersections having four-lane divided major street configuration, right turning traffic from major, i.e. movement1 is less impeded by left turns from major street. Hence, the weightage for left-turning traffic, in this case will be 0.5. Also at such intersections, due to the absence of lane discipline, right turns from minors are observed to be waiting at the intersections along the center or right side of the approach hindering the subject movement. Hence, the right turners on minor street approach, i.e. movement 7 should also be considered in computing the conflicting flow for right turn from major and is assigned a weightage of 1. The opposing major street through traffic was found to have a greater influence on right turners from major and hence a weightage of 1 is assigned to through movements on major. And for computing conflicting flow for right turn from minor, that is movement 7, the through traffic and right turn from major have greater effect on the subject movement and are assigned a weightage of 1

516

P. H. Souparnika et al.

Table 3 Conflicting flow computation Vazhuthacaud Junction Movement

Conflicting flow

Vcx (PCU/hr)

VC1

V5 + V7 + 0.5V6

1521

VC7

V5 + V1 + 0.5V2

2121

Chadayamangalam Junction Movement

Conflicting flow

Vcx (PCU/hr)

VC1

1.5V5 + V6 + V7

1850

VC7

V5 + V1 + V2

2068

Sreekaryam Junction Movement

Conflicting flow

Vcx (PCU/hr)

VC1

1.5 V5 + V6 + V7

1654

VC4

1.5V2 + V3 + V10

1563

VC7

V4 + V5 + V1 + V2

2321

VC10

V1 + V2 + V4 + V5

2321

VC8

V4 + V5 + V1 + V2 + V3 + V10

2538

and also merging through movement found a lesser effect on the subject movement, is assigned a weightage of 0.5. At the same time, when the major street is of two-lane undivided configuration and if left turning movement is not channelized, in such case, left turners from the major street will conflict with the subject movement as well, so its effect is considered. Moreover, when there is no divider on the major street, the impact of the major street through movement on the subject movement is enhanced, thus giving a weightage of 1.5. Capacity for any movement at an unsignalized intersection is computed using Equation 2 given in [1]. Table 4 shows the movement capacity at selected locations. Site-specific critical gaps for different movements were obtained using the base critical gap values. Raff’s method is used for determining base critical gap values. The critical gap for any movement is obtained using Equation 3 given in [1]. The follow-up time is obtained by taking 60% of the critical gap value [1]. Table 4 shows the movement capacity at selected locations.

Cx =

a × Vcx × e 1−e

−Vcx (tcx −b) 3600

Vcx t f x 3600

(2)

Cx = Capacity of movement ‘x’ (PCU/h) Vc, x = Conflicting flow rate corresponding to movement x (PCU/h) tc,x = Critical gap of standard passenger cars for movement ‘x’ (s) tf,x = Follow-up time for movement ‘x’ (s) ‘a’ and ‘b’ = Adjustment factors based on intersection geometry

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

517

Table 4 Movement capacity at study location Vazhuthacaud Junction Movement

tc,base

fHV

PHV

tcx

tfx

a

b

Cx (PCU/hr)

1

2.7

0.46

6.6

3.57

2.14

0.8

1.3

784

7

3.8

0.88

4.3

5.08

3.05

1.0

2.16

456

Chadayamangalam junction Movement

tc,base

fHV

PHV

tcx

tfx

a

b

Cx (PCU/hr)

1

3.5

0.78

4.2

4.62

2.77

0.7

-0.11

385

7

3.8

0.01

3.68

3.81

2.29

0.8

0.72

383

Sreekaryam junction Movement

tc,base

fHV

PHV

tcx

tfx

a

b

Cx (PCU/hr)

1

3.2

0.78

4.18

4.3

2.58

0.7

-0.11

468

4

3.2

0.78

5.7

4.5

2.7

0.7

-0.11

214

7

3.6

0.01

5.78

3.62

2.2

0.8

0.72

377

10

3.6

0.01

6.17

3.62

2.29

0.8

0.72

377

8

4.2

0.07

5.1

4.31

2.58

1.1

0.72

265

12

4.2

0.07

5.7

4.32

2.53

1.1

0.72

242

tcx = tc,base + f H V × ln(PH V )

(3)

tc, base = Base critical gap value (s) fHV = Adjustment factor for large vehicles PHV = proportion of large vehicles in the conflicting traffic stream. Determination of LOS for Movements LOS is defined as the level of perception of the service offered by a facility by the user. In the traffic engineering context, it is arrived at using measures of effectiveness such as volume/capacity ratio (v/c), delay and density Indo HCM [1] classified LOS for different v/c ratios. LOS derived based on the observed volume-capacity ratio at the intersections is shown in Table 5. LOS E for a movement implies that the movement is overloaded and operating at near capacity. It requires geometric improvements at the intersection for its stable flow. Simulation Modelling Using VISSIM VISSIM is a microscopic time step and behaviour-based traffic simulation model that efficiently replicates heterogeneous traffic flow and it is widely used today to simulate heterogeneous traffic flow. The selected intersections were modelled in VISSIM by assigning geometric features, traffic composition, vehicle volume, vehicle routes and speed, so that it represents the actual field conditions. Fig. 6 shows the modelling done in VISSIM. The developed VISSIM model is calibrated by comparing travel time values obtained from the field and VISSIM and Mean Absolute Percentage

518 Table 5. LOS calculation for movements

P. H. Souparnika et al.

Vazhuthacaud junction Movement

V/C Ratio

LOS

Condition

1

0.89

E

Unstable flow

7

0.92

E

Unstable flow

Chadayamangalam junction 1

0.86

E

Unstable flow

7

0.62

D

High-density flow

Sreekaryam junction 1

0.98

E

Unstable flow

4

0.37

C

Stable flow

7

0.5

C

Stable flow

10

0.3

B

Stable flow

8

0.72

E

Unstable flow

11

0.83

E

Unstable flow

Error (MAPE) values were found to be within allowable limit, i.e. less than 15%. The validation of the model was done by evaluating a calibrated simulation model with a new set of field data. The MAPE values were found to be within the allowable limit, i.e. less than 15%. Safety Assessment Using SSAM After developing the simulation model in VISSIM, the vehicle trajectory files were exported to SSAM software. Several indicators for traffic conflicts were computed by SSAM, involving post encroachment time (PET), time to collision (TTC), max speeding (Max S), the rate of deceleration (DR) and max deceleration (Max.D). The number and type of conflicts existing at the site were also evaluated by SSAM. The conflicts were categorized into three types according to the conflict’s angle rearend, lane change and crossing conflicts. The conflicts are classified according to the absolute value of the conflict angle. The conflict type is classified as a rear-end conflict if the conflict angle < 30°, a crossing conflict if the conflict angle >85° and a lane change conflict if 30° ≤ conflict angle ≥ 85°. The SSAM computes

Fig. 6 Simulation modelling done in VISSIM

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

519

several surrogate safety measures, among these the most important indicators, postencroachment time (PET) and time to collision (TTC) were used for analysis. Fig. 7a, 7b and 7c shows the TTC and PET variations for different conflict types at selected intersections. Table 6 shows the SSAM output results of selected intersections and Table 7 shows the comparison between crossing conflicts obtained at field and SSAM. The crossing conflicts obtained from the field and SSAM shows close approximation, thus SSAM is the best tool to perform safety analysis at the field.

4 Conclusions Conflicting flow for any movement should include all those movements of higher priority with which the subject movement competes for the right of way. However, due to the absence of movement priorities at uncontrolled intersections in India, the conflicting movements and their contributions towards conflicting flow rates will be different. Based on the site geometry and relative impact of conflicting movements on subject movement, the computation of conflicting flow rates was modified. Using these conflicting flows, the movement capacity was determined. Thus, the proposed methodology can be used for finding the capacity of movements at uncontrolled intersections functioning under heterogeneous traffic conditions. Also, the determination of the V/C ratio helps to find out the LOS for each movement and to know the prevailing conditions of movements at the site. The study helps to suggest geometric improvements required at intersections to make flow stable for different movements. The safety assessment done at selected intersections using VISSIM and SSAM computes various surrogate safety measures, classifies conflicts according to conflict angles and also determines the total number of conflicts at intersections. Lower traffic and congestion at the Chadayamangalam intersection result in a higher value of mean TTC and PET value which lies above 2.5 s. Higher traffic flow at the Sreekaryam intersection having a four-legged configuration showed a lower value of mean PET value of about 2.26 s compared to other intersections. Heavy through movements and more number of turning traffic at the Sreekaryam intersection results in greater number of conflicts compared to other intersections. The queue formation at the site by the drivers to find a suitable gap for their desired maneuvers generates a higher percentage of rear-end conflicts at all intersections. Compared to other conventional safety analysis, the use of simulation software VISSIM and SSAM consumes less time for safety evaluation. By extracting path files constructed using VISSIM, SSAM classified conflicts according to conflict angle and also determined the number and severity of conflicts at intersections.

520

P. H. Souparnika et al.

5 3.52

TTC & PET (s)

4

3.16

2.94 3 2.78 2

2.67

2.26

1

Mean PET

Mean TTC

0

Vazhuthacaud

Chadayamangalam

Sreekaryam

TTC & PET (s)

5 4

3.77

3.68

3.04 3

3.22

2.98

2.68

2 1

Mean PET

Mean TTC

0 Vazhuthacaud

Chadayamangalam

Sreekaryam

TTC & PET (s)

5 3.39

4 3

3.08

2.65

2 1.84

1.99

1 Mean TTC

Mean PET

1.18

0 Vazhuthacaud

Chadayamangalam

Sreekaryam

Fig. 7. a TTC and PET variations for crossing conflicts at selected intersections. b TTC and PET variations for lane change conflicts at selected intersections. c TTC and PET variations for rear end conflicts at selected intersections

Capacity Analysis and Safety Assessment of Unsignalized Intersection …

521

Table 6 SSAM output results of selected intersections Mean TTC (s)

Vazhuthacaud

Chadayamangalam

Sreekaryam

2.48

2.95

2.59

Mean PET (s)

2.37

2.63

2.26

Crossing conflicts

278

225

395

Lane change conflicts

133

171

235

Rear end conflicts

426

527

697

Total conflicts

837

923

1327

Table 7 Comparison of crossing conflicts obtained from field and SSAM Intersections

No. of crossing conflicts No. of crossing conflicts Percentage difference from field from SSAM

Vazhuthacaud

249

278

11.64

Chadayamangalam 198

225

13.64

Sreekaryam

395

11.58

354

References 1. Indian Highway Capacity Manual (Indo-HCM), CSIR, New Delhi (2017) 2. Babu SS, Vedagiri P (2016) Proactive safety evaluation of a multilane unsignalized intersection using surrogate measure. Transportation letters 10(2):104–112 3. Joewono P, Ning W, Kamarudin A, Mohd ES, Basil DD, Josef H (2016) Performance of nonpriority intersections under mixed traffic conditions based on conflict streams analysis. J Transp Eng, PartA: Syst 144(6):04018024 4. Pitlova E, Kocianova A (2019) Determination of priority stream volumes for capacity calculation of minor traffic streams for intersections with bending right-of-way. Transportation Research Procedia 40:875–882 5. Ghanim MS, Shaaban K (2019) A case study for surrogate safety assessment model in predicting real-life conflicts. Arab J Sci Eng 44(5):4225–4231 6. Mithun M, Satish C (2019) Capacity estimation of unsignalized intersections under heterogeneous traffic conditions. Can J Civ Eng 47(60) 7. Arkatkar SS, Chepuri A, Joshi G (2020) Developing proximal safety indicators for assessment of unsignalized intersection–a case study in Surat city. Transportation letters 12(5):303–315 8. Hussein MA, Hussein AE (2021) Coupling Visual Simulation Model (VISSIM) with Surrogate Safety Assessment Model (SSAM) to evaluate Safety at Signalized Intersections. J Phys: Conf Ser 012234 9. Goyani J, Aninda BP, Ninad G, Shriniwas A, Gaurang J (2021) Investigation of crossing conflicts by vehicle type at unsignalized T-intersections under varying roadway and traffic conditions in India. Journal of Transportation Engineering, Part A: Systems 147(2):0502001 10. Paul M, Ghosh I (2019) Post encroachment time threshold identification for right-turn related crashes at unsignalized intersections on intercity highways under mixed traffic. Intenational journal of injury control and safety promotion 27(2):121–135

A Statistical Approach to Estimate Gap Acceptance Parameter at Three-Legged Uncontrolled Intersection Khushbu Bhatt

and Jiten Shah

Abstract The estimation of the gap acceptance parameter for right-turning movements at uncontrolled T-intersections is the main focus of this paper. It takes a lot of effort to investigate an intersection that is unsignallised or is uncontrolled. In general, a driver’s perception of the priority at an intersection is based on the volume of traffic, the design of the intersection, and the speed of the vehicle on main and minor approaches. Three uncontrolled T-intersections in the field were recorded as video in order to acquire the data. Accepted gaps, rejected gaps, vehicle types, and traffic composition are the data that were extracted from the video. The accepted and rejected gaps are fitted with various statistical distributions. The bestfitted model is recommended for the accepted and rejected gap based on goodness of fit tools, and the critical gap is estimated. Additionally, the critical gap value determined by best-fitted distributions (log-normal distribution) is compared to the values determined using Indo-HCM (Indo-HCM: Indian Highway Capacity Manual (Indo-HCM). CSIR-Central Road Res. Institute, New Delhi. (2017)). The value of the gap obtained using detailed traffic parameters can be further used for risk prediction to reduce the severity of crashes at uncontrolled intersections under mixed traffic conditions. Keywords Gap acceptance · Maximum likelihood estimate · Critical gap

1 Introduction and Background In developing countries, like India, the unsignallised intersection does not follow priority rules for right-turning movement. The drivers themselves decide the acceptance or rejection of the gap which depends on the geometry, speed of vehicles, type K. Bhatt (B) Department of Civil Engineering, Drs. Kiran and Pallavi Patel Global University, Vadodara, India e-mail: [email protected] J. Shah Institute of Infrastructure Technology Research and Management, Ahmedabad, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_42

523

524

K. Bhatt and J. Shah

of conflicting vehicles, etc. Hence, the gap accepted or rejected at intersections plays a very vital role in right-turning movement at an uncontrolled intersection. The gap is the minimum space or time interval within which the vehicle will safely make a maneuver from a major street to a minor street or vice-versa. The tendency of drivers that intellects the opportunity to move further to cross the intersection safely is known as gap acceptance behavior. This concept is useful for switching the traffic stream from major to minor or even in case of lane changing. The gap acceptance parameter generally depends on the gap and lag of the vehicle in the field. However, to evaluate the behavior of the driver for acceptance or rejection of the gap further depends on many other parameters like occupancy time, conflicting vehicle type and speed, etc. The gap acceptance parameter is the micro-level traffic characteristics that explain the interaction of traffic when the minor street vehicle enters the major street. The critical gap and follow-up time are the two important parameters of gap acceptance. As per the Indian Highway Capacity Manual [1] the critical gap is defined as the minimum time between successive major-stream vehicles in which minor stream vehicles make a manoeuver. The value of the critical gap lies between the accepted and the largest rejected gap. Critical Gap is the smallest gap that a driver is willing to accept to merge with the conflicting traffic and mainly determines the gap acceptance behavior of the driver. A vehicle making the turn to either direction increases the risk of road accidents if the accepted gap is less than that of the critical gap. Moreover, the value of the critical gap is used to calculate the capacity of the intersection, delay, and level of service [2, 3]. The follow-up time also plays a crucial role to ensure the driver’s safety. The follow-up time is defined as the time between two successive vehicles from the minor street entering the conflict area using the same gap. The follow-up time is generally considered to be 0.6 times the critical gap [1]. The estimation of the critical gap is a challenging task because it cannot be directly deduced from the data obtained from field observations, it needs to be estimated based on the accepted or/and rejected gap. A minor street driver at an intersection needs to judge the gap and whether it is suitable to cross the traffic stream safely. In general, the driver accepts all gaps larger than the critical gap and rejects the other gap [4]. Hence, the critical gap is the least value of the gap that is acceptable to a driver. In the Highway Capacity Manual [1], the terminology of the critical gap is replaced with the critical headway, however, the universal applicability of the term has not been justified. The critical gap is not a constant value but varies on individual perceptions of drivers. The various parameter that affects the value is a vehicle type, the geometry of intersection, delay, gradient, etc. [5–8] but the value is independent of conflicting traffic volume [9, 10]. As critical gaps cannot be measured in the field, continuous efforts are made by researchers to develop new techniques to estimate the mean critical gap. Therefore, different methods are recognized for estimating the critical gap which broadly classifies as a deterministic, probabilistic, stochastic, and fuzzy model. A deterministic method is a conventional approach that provides a single average value whereas the probabilistic method solves the inconsistency elements in gap acceptance behavior by using a statistical approach [11]. From the past literature, few studies have been claimed that

A Statistical Approach to Estimate Gap Acceptance Parameter …

525

the maximum likelihood method estimates the precise results as compared to other methods like the Raff’s method underestimates the value whereas the Ashworth method overestimates the value of the critical gap. Hence in this paper, MLE is focused on considering different distributions of the accepted and rejected gap. The use of the different distribution is to understand the realistic behavior of the driver on the section. The present paper aims to estimate the critical gap by applying the maximum likelihood estimation (MLE) process by ten different statistical distributions. Further, to check the suitability of the distribution, the gap acceptance data are analyzed by the goodness of fit. The best-fitted model for the critical gap is chosen based on the Akaike Information Criterion (AIC). Several models and distributions have been developed by various researchers, based on the traffic situation, capacity, field conditions, vehicle specifications, human behaviour, and other elements for developed and developing countries. The comparative analysis of the various models for estimating the critical gap by data required, methodology, and its limitations [12]. Based on the assumptions and limitations of the different methods for estimation of the critical gap, a simple technique of MLE is proposed in this paper by understanding the realistic behavior of drivers at the intersection.

2 Approach for Estimating Driving Gap Acceptance Behavior As different researchers have suggested, the maximum likelihood method is the best when compared with the other methods. But, it has the assumption of the driver to be constant and homogeneous which is practically impossible, especially, in the case of the heterogeneous traffic condition [13]. Hence, the approach of maximum likelihood is used in this paper with consideration of the individual driving behavior. The gap acceptance behavior of the driver is being observed using the different distributions and the best-fit model is used for the estimate of the value of the critical gap.

2.1 Maximum Likelihood Estimation In this section, the methodology of the Maximum Likelihood Estimation (MLE) [14] procedures for fitting accepted and rejected gaps and critical gaps is discussed in detail. The MLE is one of the popular and widely used methods of estimation and it holds prominent properties such as consistency, asymptotic normal, and invariance. Suppose X and Y denote the accepted and rejected gap time, respectively, and follow the same distribution function with probability distribution function (pdf), f X (x, θ )(or f Y (y, θ )) with unknown parameter θ . The likelihood function for the accepted gap time X is defined by.

526

K. Bhatt and J. Shah n ( ) | | θ |x = f X (xi , θ )

(1)

i=1

where x = {x1 , x2 , . . . , xn } the observed sample is on accepted gap time X and n denotes the number of drivers. Similarly, the likelihood function Y can be defined by n ( ) | | θ |y = f Y (yi , θ ),

(2)

i=1

where y = {y1 , y2 , . . . , yn } is the observed sample on rejected gap time Y. For the given data, the MLE estimate of the unknown parameter θ is the value that maximizes the likelihood or log-likelihood function. It illustrates the procedure for Weibull distribution and other distributions that can be implemented similarly. The pdf of the Weibull distribution is given by α

f W (x) = αθ x α−1 e−θ x , θ > 0, α > 0, x > 0.

(3)

The log-likelihood function for the Weibull distribution is given by n n E E ) ( log(xi ) − θ xiα . Log| θ, α|x = nlog(θ ) + nlog(α) + (α − 1) i=1

(4)

i=1

The MLEs of α and θ are the numerical values, at which, the Log|(.) function achieves its maximum. The usual optimization procedure gives /

θ mle = E n

n /

α mle i=1 x i

(5)

/

where α mle is the solution of the non-linear equation n n E n E n + log(xi ) − En xiα log(xi ) = 0. α α x i=1 i i=1 i=1

(6)

The equation above can be numerically solved using Newton’s method. Further, the MLE procedure for estimating the critical gap parameter using the accepted and rejected gap times given for each driver. The procedure is based on the fact that critical gap time cannot be directly observed on the field but it is known that its value (say Z) lies in between accepted and rejected gap times with probability P[Y < Z < X ] = F(X ) − F(Y ), where F(.) denotes the distribution function assumed for the critical gap variable. The assumption has been made for the Z function, as it follows the Weibull distribution. The likelihood function is given by

A Statistical Approach to Estimate Gap Acceptance Parameter …

527

n n ) | ( | | −θ yi α α| e | θ, α|x, y = − e−θ xi , [F(xi ) − F(yi )] = i=1

(7)

i=1

where xi (yi ) denotes rejected and accepted gap time for i th driver. The MLEs of α and θ can be obtained by solving the following log-likelihood equations ) ( ∂Logl θ, α|x− , y −

∂θ

|

) ( ∂Logl θ, α|x− , y

| −θ y α

α n E xiα e−θ xi − yiα e i | −θ y α = 0, α| ∂α e i − e−θ xi i=1 | | α α n E xiα log(xi )e−θ xi − yiα log(yi )e−θ yi | −θ y α = =0 α| e i − e−θ xi i=1

=

−

(8)

The equations given above need to be solved numerically using Newton’s method. Once the MLEs θ mle and α mle are obtained, the MLE of the critical gap (mean), variance, and median are given by /

/

) ( | 1 + 1/α , C R mle = 1/α θ { ( ( / ) / )} = θˆ −2/αˆ | 1 + 2 αˆ − | 2 1 + 1 αˆ , /

/

/

/

/

St D mle

−1/α | /

/

/

Md mle = θ

log2

|1/α

(9) (10)

/

.

(11)

A similar procedure can be applied to perform estimation for other probability distributions. After the MLE, the values of AIC are used to measure the goodnessof-fit for the accepted gap, rejected gap, and critical gap. The Akaike Information Criterion (AIC) is a statistical measure used for the selection of the best possible model for the given data set. The AIC has two components; one is based on likelihood and the other includes the number of parameters to be estimated through the data. The later part penalizes the model for having a higher number of parameters that increase the computational costs. Hence, AIC is applied in this study to investigate the risk of underfitting and overfitting of the data. A smaller AIC value corresponds to a better-fitting model. It is mathematically represented as. AIC = 2k − 2 ln (),

(12)

where k is the number of estimated parameters in the model; L´ is the maximum of the likelihood function for the model.

528

K. Bhatt and J. Shah

3 Data Collection The survey was conducted for three locations to analyse the gap acceptance behaviour of Indian drivers on similar geometric sections. The data collection had been done for different vehicle types at uncontrolled T- Intersection of state highways. The videography for classified traffic volume count was conducted for 12 h from 9:00 AM to 9:00 PM on working days in fair weather conditions. The accepted and the rejected gaps were calculated from the video for one peak hour in the morning and evening. Presently, three rural intersections of Gujarat are considered Nadiad (SH 60)–L1, Anand (SH 83)–L-2, and Halol (SH 5)-L3. All three selected intersections are four-lane divided–major roads and two-lane undivided- minor roads. The locations are shown in Fig. 1. The videography was done by placing a camera at a vantage point to obtain precise data as per the standard guidelines. The geometric and traffic details of all the sections are represented in Table 1.

Fig. 1 Study location as T- intersection a Nadiad (L1) b Anand (L2) c Halol (L3)

Table 1 Details of the study locations Location

Nadiad (SH 60)-L1

Anand (SH 83)-L2

Kalol (SH 193)-L3

Geometry

Three-legged Intersection

Three-legged Intersection

Three-legged Intersection

Section–Major Road Minor Road

Four-lane divided Two-lane undivided

Four-lane divided Two-lane undivided

Four-lane divided Two-lane undivided

Height of Camera

15 m

13 m

12 m

Width of Major road

15 m

15 m

17.5 m

Width of Minor road

10.5 m

10.5 m

7.5 m

Conflict area

16 m*20 m

16.6*14 m

18*20 m

A Statistical Approach to Estimate Gap Acceptance Parameter …

529

4 Data Extraction The traffic volume is extracted manually from the recorded video for classified vehicles. Subsequently, the traffic counts are converted into equivalent passenger cars; using PCU values provided by Indo-HCM 2017. The data extracted denotes the major proportion of two-wheeler traffic at all intersections and due to the smaller area of vehicles, the motorized vehicles have more aggressive behavior at the intersection during maneuver [15]. Therefore, the critical gap for two-wheelers is analyzed for all three intersections in this paper. The values are estimated using various statistical models using the maximum likelihood approach and then are compared with the other methods.

4.1 Accepted and Rejected Gap Data for the measurement of accepted gaps is extracted for each possible movement. For understanding, the movement from the major street to the right turn is presented in Fig. 2. When a vehicle has free movement from a major to a minor-street or viceversa, it is considered as an accepted gap but when the vehicle applies brake in the conflict area for a particular time during the movement is considered as a rejected gap. Figure 2 illustrates that when a vehicle is moving from point A to B, without applying any brake in between is termed as the accepted gap. It simply means that the driver has accepted the gap for crossing the manoeuvre. A vehicle with random movement is shown in the figure with a blue line moving from a major street (1) to a minor street (3) with a right turning, the time required to travel from A to B is the accepted gap, and time required to travel from A to C is occupancy time. Whereas, the green shaded area is the conflict area of the intersection. In the present study, to understand the effect of gap acceptance behaviour of the driver at an uncontrolled intersection for right turning movement, the data size of 60, 100, and 86 observations for two-wheelers are considered for L1, L2, and L3,

Fig. 2 Measurement of an accepted and rejected gap

530

K. Bhatt and J. Shah

respectively. It depends on the traffic flow for one hour, as in the case of L2 and the proportion of vehicles was more as compared to location 1. The descriptive statistic for the values of the accepted gap and the rejected gap is shown in Table 2. The minimum value of the accepted gap is 1.09 s and the maximum value is noted to be 8.06 s, whereas the minimum rejected gap is 1.85 s and the maximum is 9.54 s. The obtained value of Pearson’s coefficient of skewness (CoS) [16] for the data on the accepted and rejected gap for all three locations is designated. For the accepted gap, the CoS for L1, L2, and L3 are 0.724, 0.185, and 0.277, respectively. For the rejected gap, the CoS for L1, L2, and L3 is 0.349, 1.392, and 0.485, respectively. These CoS indicate that the empirical distribution is positively skewed having tailed at right in all six cases. It indicates that the value of the accepted gap is more towards the positive side which results in enhancing the prediction accuracy of the critical gap using distributions. In this section, simple statistical distributions are considered to be competing with each other and used for fitting time-to-event data. They are called the Weibull, Log-normal, Chen, Generalized Exponential, Generalized Lindley, Gamma, Burr, Gompertz, Power Lindley, and Lomax distributions. Each distribution is indexed by two parameters and is pertinent to use in practice. There is a lack of research in the existing literature related to the best-suited statistical distribution for the data sets using the goodness-of-fit criterion. First, fit accepted and rejected gap times individually by all ten distributions. Later, the estimation of the critical gap value using accepted and rejected gap values was combined in the MLE procedure. Table 3 shows AIC values to understand how the model fits the data sets without over or underfitting it. The AIC value models that achieve high goodness of fit and deal severely with them become complex. AIC score is of much use when compared the score with the competing model. The lower the AIC score the better the model is suitable and this shows the balance between its ability to fit the data. Therefore, as per the least value (deviation) of AIC the accepted and rejected gap defines the driver’s realistic behaviour on the field. For the accepted gap, the lognormal and generalized exponential is the best-fitted distribution for all three locations. Whereas, rejected gap follows the gamma and burr for location 1; Weibull and Burr for location 2 and location 3, respectively. The distribution is substantiated by determining the Table 2 Statistics for accepted and rejected gap (sec) for two-wheelers Parameters

Accepted gap

Rejected gap

L1

L2

L3

L1

L2

L3

Minimum

1.12

1.09

1.46

1.85

1.85

2.23

Maximum

5.90

8.06

7.72

8.56

9.16

9.54

Range

4.78

6.97

6.26

6.71

7.31

7.31

Mean

2.55

2.94

3.89

4.68

5.99

5.77

Median

2.27

2.85

3.71

4.46

5.15

5.45

Std. Dev

1.16

1.46

1.95

1.89

1.81

1.98

A Statistical Approach to Estimate Gap Acceptance Parameter …

531

goodness of fit. The least AIC value is considered for the distribution fit as very little variation in the value of log-normal and G-exponential for an accepted gap for all three locations. The Kolmogorov–Smirnov (K-S) test and Anderson–Darling (A-D) test values are particularized to validate the best fit for the gap acceptance parameter in Table 4. In Table 4, validation is done for the four distributions having minor variations in the AIC values for the accepted gap. Using easy-fit software, distribution fitting is executed which as a result obtains the p-value of the two tests named the K-S test and the Anderson–Darling test. This p-value assists in identifying the appropriate distribution whichever has a lesser p-value. As per the K-S test and Anderson–Darling test, the least p-value is obtained for the log-normal distribution as shown in Table 6. This distribution shows the realistic gap acceptance behavior of driver on the field which enable to attainment of the specific value of the critical gap. The goodnessof-fit result verified that the log-normal is best suitable for the accepted gap which will be further used to estimate the value of the critical gap using MLE.

4.2 Different Statistical Distribution for Estimation of Critical Gap By different distribution, this paper tends to understand the realistic behaviour of drivers which is inconsistent depending on several parameters like vehicle type, traffic and geometrical characteristics, driver behaviour, speed, etc. The MLE is used as a statistical method of estimating the parameters of probability distribution by maximizing the likelihood function so that in the statistical model the observed data depicts the risk-compelling behavior of the driver on the field which enables to predict the proximate value of the critical gap. A driver’s behaviour in a particular location is different and the gap value also varies based on the geometrical features of the intersection. To observe the difference, the accepted and rejected gap of the two-wheeler is considered in this paper for three intersections having similar road features (uncontrolled T-intersection). The different statistical distribution functions are applied using R-software by the MLE method. All ten models’ cumulative distribution functions are plotted for three locations to select the best fit out of all. The models are selected based on their suitability for two-parameter estimation. The input parameters are accepted and rejected gaps for two-wheelers. The cumulative distribution function of the parameters is represented in Fig. 3 for L1, L2, and L3, respectively. It can be seen from the graph that the value of the critical gap (shown in blue colour) lies between the accepted and rejected gap. In Fig. 3a, for location 1, the MLE method is applied for ten likely distributions to get the best fit and obtain the precise value of the critical gap. Weibull distribution, the deviation for the accepted gap is between 2 and 4 s whereas for the rejected gap deviation is observed in the range of 3−8 s. In a similar way when all the curves are observed with the actual values and deviation from the distribution curve the

116.840

Rejected Gap

194.680

Rejected Gap

160.779

174.718

Accepted Gap

Rejected Gap

Location 3

158.692

Accepted Gap

Location 2

85.259

Weibull

Accepted Gap

Location 1

176.146

156.802

196.081

148.919

116.403

79.650

Gamma

178.482

154.603

198.729

143.982

117.121

77.038

Log-Normal

178.573

155.446

198.534

144.670

117.000

77.155

G-Exponential

Table 3 AIC for statistical distributions for accepted and rejected gap

175.429

159.695

195.250

155.374

116.491

82.549

Burr

177.982

170.103

199.226

177.341

120.448

95.379

Chen

178.522

169.700

199.840

176.478

120.690

95.279

Gompertz

175.780

160.633

195.359

157.513

116.897

84.751

P-Lindley

177.874

156.245

197.814

145.740

116.859

77.700

G-Lindley

232.771

193.017

256.888

197.226

148.611

112.244

Lomax

532 K. Bhatt and J. Shah

A Statistical Approach to Estimate Gap Acceptance Parameter …

533

Table 4 Goodness-of-fit for statistical distributions of accepted gap Log-Normal

Gamma

Weibull

G-Exponential

0.0960

0.1173

0.1207

0.1608

L-1

K-S Test A-D Test

0.2718

0.5640

1.1236

1.7916

L-2

K-S Test

0.1014

0.1168

0.1424

0.1699

A-D Test

0.6343

0.8012

1.5484

2.2479

L-3

K-S Test

0.1215

0.1216

0.1346

0.1943

A-D Test

0.6981

0.7111

1.1867

1.1425

minimum deviation is observed for the lognormal distribution for the accepted gap. However, the maximum deviation from the observed curve from the actual value for the accepted and rejected gap is perceived in the Lomax distribution. The deviation in the values in the graph is statistically represented using the AIC value in Table 5 to get the best fit for the value of the critical gap. In Fig. 3b, for location 2, the deviation of the curve for Chen, Gompertz, Lomax and p-Lindley distribution is higher as compared to the other distribution. In the Weibull distribution, the deviation is observed for the accepted gap in the range of 3–4 s whereas no observable deviation in the case of a rejected gap. Similarly, for G-exponential and lognormal distribution, the difference in the observed value and curve varies slightly which is estimated using the AIC value of the critical gap depicted in Table 5. The value of AIC is 81.844 and 81.988 for G- exponential and lognormal distribution. The exponential distribution is based on the calculation of the product of reliability and cannot be used to assume future probabilities from the records. Therefore, the lognormal distribution will be considered for hazard function analysis in further research. In Fig. 3c, for location 3, the deviation of the curve for accepted and rejected gap from the observed values is depicted for all the ten distributions in which Weibull distribution shows deviation ranging between 2 and 4 s in the accepted gap whereas for rejected gap less variation is observed in the range of 4−6 s. In the case of gamma, lognormal, and burr distribution slight deviation for the accepted and the rejected gap is observed whose statistical measure is depicted using AIC value for the critical gap. The AIC value for lognormal and gamma distribution is 125.440 and 125.879, respectively. Although gamma and lognormal have similar curves when its log value is considered the gamma has a heavier tail on the negative side and is negatively skewed which results in less value on the left side of the curve. But in the field, values of the accepted gap range maximum proportion in 1.09−5.9 s (Table 2), and if the curve is negatively skewed the value will be more on the right side, i.e. on the higher range which will vary the result. Therefore, log normal will be considered to get the precise value of the critical gap. For the Lomax distribution, the deviation is slightly higher for both accepted and rejected gaps, hence this distribution is neglected and cannot be considered for further research or analysis. As per the least AIC values of the critical gap, the best-fitted model is identified for three locations. In general, the best fit is either given by Generalized Exponential

534

K. Bhatt and J. Shah

Fig. 3 Maximum likelihood estimates cumulative distribution function different locations a L–1 b L–2 c L–3

A Statistical Approach to Estimate Gap Acceptance Parameter …

Fig. 3 (continued)

535

536

Fig. 3 (continued)

K. Bhatt and J. Shah

L3

L2

L1

3.409

3.382

1.227

Median

Std Dev

1.724

Std Dev

1.714

4.403

4.611

4.654

4.607

Mean

125.440

127.057

Median

AIC

3.544

3.697

1.464

Median

Std Dev

1.326

3.705

129.304

1.138

3.230

3.356

82.755

Gamma

3.756

Mean

135.981

85.854

AIC

Mean

AIC

Weibull

Model

1.825

4.279

4.601

125.879

1.353

3.439

3.665

127.301

1.166

3.147

3.333

81.948

Log-Normal

1.839

4.279

4.603

125.936

1.333

3.418

3.651

127.107

1.163

3.120

3.320

81.844

G-Exponential

1.677

4.488

4.631

125.988

1.298

3.623

3.731

131.253

1.113

3.279

3.372

83.468

Burr

1.424

4.787

4.160

133.053

1.341

3.819

3.501

148.615

1.228

3.522

3.297

91.826

Chen

1.938

4.798

4.629

133.666

1.754

3.809

3.730

149.713

1.431

3.532

3.406

92.149

Gompertz

Table 5 Estimated AIC, mean, median, and standard deviation for the distribution for all three locations

1.700

4.550

4.642

126.567

1.405

3.670

3.741

134.090

1.189

3.348

3.394

85.148

P-Lindley

1.801

4.314

4.607

126.790

1.311

3.443

3.654

127.461

1.145

3.139

3.324

81.984

G-Lindley

4.585

3.174

4.577

175.293

3.719

2.573

3.710

189.412

3.436

2.380

3.433

120.962

Lomax

A Statistical Approach to Estimate Gap Acceptance Parameter … 537

538

K. Bhatt and J. Shah

or Lognormal for locations 1 and 2. The mean value for the critical gap for the best-fit model is 3.32, and 3.33 s for L-1. And the mean value for the critical gap estimated is 3.65 and 3.66 s for L-3. For location 3, the best-fit models are Gamma and lognormal distributions and the value of the critical gap is 4.60 and 4.61 s, respectively. The value of the critical gap varies marginally but for the safety analysis, a few seconds are also precise to reduce the severity of accidents. A minor error in the estimated value may lead to a severe crash. Hence, the best distribution is required for evaluating the gap acceptance parameter for each location. The log-normal distribution is best suited for all three locations and will be considered for further analysis of risk and severity. It is well-defined that the value of the critical gap is independent of conflicting traffic volume [9, 10] nevertheless this proved incorrect; traffic volume and speed indirectly affect the value of the critical gap.

5 Results and Discussion The critical gap has been estimated using MLE with lognormal distribution and selected on the basis of AIC value as it is one of the suitable statistical analyses in order to choose the best-fit model. Hence the least AIC value is considered to be the quality model as it loses fewer input data which minimizes the error. As the estimated value of the critical gap is based on the input parameter of the accepted and rejected gap, therefore for further check, the goodness-of-fit for the gap acceptance parameter is applied for validation. Five out of ten distribution functions are found to be the most suitable model. To be more precise, the different models are used to select the best-fit model out of five by comparative analysis and therefore, two statistical tests were performed, i.e. K-S test and Anderson–Darling test for further research. Out of which log-normal shows the best suit for each location for heterogeneous traffic conditions of rural highways. The critical gap estimation using Indo-HCM cannot be used for the further analysis of the safety and severity prediction as it considers parameters of the traffic and geometric features. Whereas for the implication to safety, the other parameters like vehicle type, conflicting speed, gap acceptance behavior of driver are required for accurate prediction of safety at the intersection. Even various statistical techniques are used for evaluating the gap acceptance parameter value which depends only on the driver behavior rather than the other parameters like conflicting vehicle, speed, gap acceptance behavior of driver with vehicle type. The distribution enables us to know the realistic behavior of the driver at the intersection. But it needs to be specific when it is been analyzed for risk and severity prediction, the precise value of critical gap in such kind of analyses add a great value for safety at intersection depending on different geometric and traffic characteristics. A suitable distribution for the gap acceptance behavior enables to obtain the precise model to reduce the severity of accidents at an uncontrolled intersection. Hence, the MLE technique is found to be more suitable as discussed in the previous sections

A Statistical Approach to Estimate Gap Acceptance Parameter …

539

and it can be used to get the critical gap and use it for risk and severity analysis for uncontrolled intersection.

6 Conclusion Based on the past literature, different statistical techniques are developed for the estimation of critical gaps to estimate the capacity and LOS of the intersection. However the researchers have lacked their emphasis on the implication of critical gap to safety which is a point of major concern at unsignallized intersections. Hence estimating the precise value of the critical gap is important Hence, the maximum likelihood estimate with the lognormal distribution provides the best estimates of the critical gap parameters. The use of appropriate distribution will enable to prediction of the suitable value and can be used for further analysis of safety prediction. This paper reveals that taking different distributions for estimating the value of critical gaps is feasible depending upon the distribution fit for gap acceptance and rejection. The value of the critical gap may vary with smaller variation and it is important to identify the best distribution among all. Hence, AIC technique was selected to identify the best fitting model by analyzing the gap acceptance parameter i.e. accepted and rejected gap which is difficult to analyze by any other statistical analysis. Therefore, it is recommended to consider various probability models and choose the best-fitting model using the AIC. Hence, it can be enlightened from the results obtained that the driver behavior is adequate to estimate the value of the critical gap. The impact of the proportion of heavy vehicles and traffic volume indirectly impact the value obtained from the statistical technique of MLE. Some of the conclusions are drawn based on the analysis: . The accepted gap rejected gap, and critical gap follows different distributions based on the geometric and traffic characteristics of the section. But lognormal is the best-suited distribution for three-legged intersections under heterogeneous traffic conditions for rural highways. . The critical gap obtained by the maximum likelihood estimate using different distributions is similar to the values attained from the Indo-HCM 2017. It demonstrates that the critical gap value estimated using the driver’s gap acceptance behavior is similar to the value calculated using Indo-HCM. Hence, the gap acceptance behavior is influenced by the traffic volume and proportion of heavy vehicles which results in the estimating the similar value when different parameters and technique is used. . As the critical gap at the uncontrolled intersection indicates the gap acceptance behavior of a driver, the precise value needs to be obtained using traffic, geometric, and gap acceptance parameters which can be used for further analysis of risk and severity of the driver at uncontrolled intersection.

540

K. Bhatt and J. Shah

References 1. Indo-HCM: Indian Highway Capacity Manual (Indo-HCM) (2017) CSIR-Central Road Res. Institute, New Delhi 2. Patnaik AK, Krishna Y, Rao S, Bhuyan PK (2017) Development of roundabout entry capacity model using INAGA method for heterogeneous traffic flow conditions. Arab J Sci Eng 42:4181– 4199. https://doi.org/10.1007/s13369-017-2677-x 3. Patkar M, Dhamaniya A (2020) Developing capacity reduction factors for curbside bus stops under heterogeneous traffic conditions. Arab J Sci Eng 45:3921–3935. https://doi.org/10.1007/ s13369-019-04309-4 4. Mohan M, Chandra S (2016) Assessment techniques for estimating critical gap at two-way stop-controlled intersections. pp 1–18 5. Aerde MV, Agar SY (1984) Capacity, speed, and platooning vehicle equivalents for two-lane rural highways. Transp Res Rec 971:58–67 6. Tian Z, Troutbeck R, Kyte M (2000) A further investigation on critical gap and follow-up time. In: 4th Int Symp Highw Capacit, pp 397–408 7. Kareem YAA (2010) A comparative study of gap acceptance at priority intersections. In: Presented at the conference 8. Rakha H, Sadek S, Zohdy I (2011) Modeling differences in driver left-turn gap acceptance behavior using bayesian and bootstrap approaches. Procedia–Soc Behav Sci 16:739–750. https://doi.org/10.1016/j.sbspro.2011.04.493 9. Brilon W, Koenig R, Troutbeck RJ (1999) Useful estimation procedures for critical gaps. Transp Res Part A Policy Pract 33:161–186. https://doi.org/10.1016/s0965-8564(98)00048-2 10. Troutbeck RJ, Brilon W (2001) Unsignalized intersection theory. Traffic-Flow Theory 8–1–8– 47 (2001) 11. Gattis JL, Low ST (1999) Gap acceptance at a typical stop-controlled intersections. J Transp Eng 125:201–207 12. Bhatt K, Shah J (2022) An impact of gap acceptance on road safety : A critical systematic review. 7:6–22 (2022). https://doi.org/10.14254/jsdtl.2022.7-1.1 13. Pan A, Zhang X, Nakamura H, Alhajyaseen W (2020) Investigating the efficiency and safety of signalized intersections under mixed flow conditions of autonomous and human-driven vehicles. Arab J Sci Eng 45:8607–8618. https://doi.org/10.1007/s13369-020-04810-1 14. Miura K (2011) An Introduction to maximum likelihood estimation and information geometry. Interdiscip Inf Sci 17:155–174. https://doi.org/10.4036/iis.2011.155 15. Sangole JP, Patil GR, Patare PS (2011) Modelling gap acceptance behavior of two-wheelers at uncontrolled intersection using neuro-fuzzy. Procedia–Soc Behav Sci 20:927–941. https://doi. org/10.1016/j.sbspro.2011.08.101 16. Bendel RB, Higgins SS, Teberg JE, Pyke DA (1989) Comparison of skewness coefficient, coefficient of variation, and Gini coefficient as inequality measures within populations. Oecologia 78:394–400. https://doi.org/10.1007/BF00379115

Identification of Infrastructural Causative Factors for Road Accidents on Urban Arterial Roads: A Case Study of Ahmedabad Poojan Pasawala and Bhavin Shah

Abstract Every year approximately a 1.35million people lose their lives in a road accident, which is the world’s eighth leading cause of death. There are 1.51 lakh fatalities per year in India because of road accidents, making India first across the world in terms of deaths. Accidents are caused by three factors: humans, vehicles, and infrastructure. Because humans are involved in driving, human error is generally considered to blame. However, most of the time, human error and a combination of other factors are to blame, so it is not only the driver who must accept responsibility for the accident. This research aims to identify the infrastructural element of the urban arterial road of Ahmedabad that can lead to the accident by conducting a Road Safety Audit and establishing its correlation with the past three years’ accident data. The research is validated with a closed-ended questionnaire survey. From the results, the stretches are prioritized for corrective measures. The result shows that the factors related to visibility and intersection; pedestrian and cyclist; speed and road cross-section and alignment are correlated with the accident. Keywords Road safety · Road safety audit · Transportation

1 Introduction 1.1 Background Transportation is the backbone of the country’s development; it provides access and mobility to the public. In the past decade, with the increase in the population and urbanization, combined with the increase in income, no. of vehicles on the road has P. Pasawala (B) CEPT University, Ahmedabad 380009, India e-mail: [email protected] B. Shah The World Bank, New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_43

541

542

P. Pasawala and B. Shah

skyrocketed. With the increase in vehicles on the road, road users suffer accidents, pollution, and congestion; among all the three consequences, accident affects road users most negatively. Road accident is an increasing cause of death worldwide, and it has emerged to be the primary concern in the past decade. Despite all the progress in road safety, the death because of road accidents in the world is around 1.35 million, which is approximately 3700 deaths/day, which is unacceptably high [1]. Also, it is one of the leading causes of disability; approximately 20–50 million people suffer from road accidents [2]. Road accidents have increased from the 16th to the 8th cause of mortality globally between 1990 and 2016, and 54% of those fatalities are either pedestrians, cyclists, or motorcyclists. [1]. According to WHO, it is the leading cause of death for people in the 5–29 age group [1]. According to the estimates, road accidents cost the country approximately 3% of its GDP, affecting the low- and middle-income group countries to a large extent as they contribute to 93% of road accidents, with only 60% of the road vehicles in the world [3]. With roughly 17.5% of the global population, India ranks third globally for the most accidents per year. As not all incidents are reported to the police, it is believed that the number is understated by a factor of 5 to 20 times. [4]. From 1970 to 2019, the number of fatalities due to road accidents increased 12 times in India. Due to road accidents in 2019, there were 1.51 lakh fatalities per year in India, which makes it first across the world in terms of death due to road accidents, constituting 11% of the deaths worldwide due to road accidents. Because of road accidents, India is losing its people and negatively affecting the economy; the accidents account for 0.77% of the nation’s GDP. [5].

1.2 Need for Study Three factors contribute to road accidents: human error, vehicle error, and road infrastructure error. Even though accidents can be a combination of more than one error, the causation is generally pointed to only one factor–human error. This leads to an overrepresentation of accidents because of human error. There is repeated wisdom, especially for Ahmedabad city, that “Driver causes 90% of the accidents” [6], even though infrastructure-related and mechanical problems are inseparable parts of the accidents. Also, the Penal section under which police register a complaint is related to road users; hence, they will only see the driver’s error in most cases. Hence it becomes necessary to identify the elements that can probably cause the accidents and to prioritize the stretches of road for corrective measures (Fig. 1).

1.3 Scope of Study The study will be limited to the Ahmedabad area and to the Arterial roads of Ahmedabad.

Identification of Infrastructural Causative Factors for Road Accidents …

543

Fig. 1 Responsible factor for road accidents [6]

Vehicle: 2% 4%

1% 1%

Infrastructure: 3% 41%

Human: 46%

The scope will be limited to identification, correlation and to prioritize the road section and will not include the remedial measure for the identified stretch.

2 Literature Review According to a review of other research on the effect of geometric design on road safety, shorter or no sight distance and small curve radius significantly impact road safety and can significantly increase the accident rate. The reviewer also pointed to the fact that accident is a factor for many reasons, and in the geometric design also there is a significant effect of one element on the other element [7]. A study conducted in China with the help of the Bayesian network studied the combination of the factors that can affect road accidents. The results show that bad weather (which causes slippery roads) and speeding can be critical factors while single-vehicle traffic is considered. When the multi vehicular traffic is considered, then the night-time with the road not physically separated, poor braking, and the use of the wrong lane can be critical [8]. A road safety audit (RSA) was conducted on NH-58 for a 65 km stretch; with the help of the multiple-regression model, the author tried to establish the safety influencing parameter, which, if corrected, can decrease the likelihood of the accident rate. They found that the traffic volume, road marking, condition of the shoulder, cross drainage, and other warning signs significantly affect road safety [9]. Also, a similar study was conducted on the NH-4 highway with the help of a multiple linear regression model; they found that the horizontal curvature and the number of the junction on the road are related to the accident on the highway, whereas the sight distance and vertical curvature are not or rarely related to the accident. Additionally, there is a negative correlation between the number of accidents and road width and roughness. [10]. An analysis of a 360 km highway in Iran using the analytical hierarchy process concluded that the main factors affecting road safety on the selected stretch are high speeds and inappropriate or non-standard horizontal curves [11].

544

P. Pasawala and B. Shah

Ahmedabad’s urban arterial roads are the subject of the study. Various research articles were reviewed to determine the road to be used and the appropriate time to conduct the road safety audit. In Ahmedabad, accidents happen highest in peak hours, from 9 to 11 am and from 6 to 8 pm [12, 13]. In Ahmedabad, the number of accident cases has increased on urban highways like the S.G. highway, 132ft ring road, and the Narol-Naroda Highway, finding of the study states that almost 85% of a fatal accidents are found along that stretch. In Ahmedabad, the eastern side of the city has more share of the accident than other parts of the city [13]. Many low-income and middle-income households are located in this area whose primary mode of transport is either 2wheeler or walking, and research shows that the 2-wheeler and pedestrian account for 74% of the total fatalities in Ahmedabad. [13].

3 Methodology Firstly, the need for the study was identified, and the existing literature was reviewed to understand the different methods available for a road safety audit, road classification and their characteristics, and other factors related to the accident which were specific to Ahmedabad city, were also studied. In the next step, primary data was collected by conducting the road safety audit, questionnaire survey, and secondary data–the past three years’ accidents recorded by the police, was collected from Ahmedabad police. The next step is data analysis, in which Pearson’s correlation was established, and the result was validated with the expert’s opinion collected from the questionnaire survey. The road section was prioritized from the response to the questionnaire survey with the help of the Relative Importance Index.

4 Data Collection and Data Analysis 4.1 Selection of the Road From the Indian Road Congress IRC:86, 2018, and MoHUA guidelines, the criteria for the selection of the road were formed. The road section selected for the study is Chandranagar BRTS–Kankaria–Jashodanagar Cross road–Naroda Patiya; the entire road section is 17 km long (Table 1 and Fig. 2)

Identification of Infrastructural Causative Factors for Road Accidents …

545

Table 1 Characteristics of the selected road Head

Parameter

Characteristics of selected stretch

Classification of the road

Arterial road

The maximum stretch is Arterial Road

Dominant user

2-wheeler, cyclist and pedestrian, 41% cycle, 35% 2-wheeler, and car 22% other, and 2% car

Existing public transport

AMTS, BRTS, preferably railway All 3 modes are there in the station selected stretch

Land use in proximity

Mixed–use–Industrial, commercial, and residential

Industrial and residential dominant, commercial to a certain extent

Dominant income group

EWS, LIG, MIG

42% LIG, 27%EWS, 22% MIG, 9% HIG

Width of the road

The majority of the road shall be between 45 and 60 m but can be relaxed till 36-60 m

60mtr from Jashodanagar crossroad to Naroda Patiya 36 m for the majority of the section from Chandranagar to Jashodanagar crossroad

Lanes

Shall be 4 or 6 lanes (Total of both carriageways)

6 on the Narol-Naroda highway & 4 on the remaining stretch

Construction

Shall be free from ongoing major construction

Length of the road

> 12 km

Total: 17.0 km

Actual to design PCU ratio

> 0.9

Vasna bridge = 5585/6000 = 0.93 Krishnanagar (Naroda) = 5531/6000 = 0.92

4.2 Checklist for the Road Safety Audit The road safety audit checklist is prepared with reference from Indian standards IRC: SP-88: 2010, guidelines of MoHUA [14], and international standard Austroad: AGRS06-09. The road safety checklist was further divided into the following heads, and the observations were recorded in the following heads: (1) (2) (3) (4) (5) (6) (7) (8)

Visibility and Intersection Pedestrian and Cyclist Speed limit Sign, pavement, and delineation Cross-section and alignment Road-side hazard Light and Night-time issues Miscellaneous.

546

P. Pasawala and B. Shah

Fig. 2 Selected road section

4.3 Observation from Road Safety Audit For the road safety audit, the road was divided into equal sections of 500 m, except Sect. 1 of 1 km, as there is a bridge in Sect. 1, so there were 33 sections on the selected road. For RSA, the road section was divided into 500 m stretches considering the spacing of major intersections on the selected stretch and BRTS stands, as the pedestrian will be crossing the street more frequently near the BRTS stand. If the road sections are divided into shorter segments, some stretched will have only midblocks, and if divided into longer segments, more than two major intersections are coming, leading to less accurate RSA data. The RSA were conducted two times a day for a particular stretch; during morning peak hours and evening peak hours, which are from 9 to 11 am and from 6 to 8 pm, as identified in the literature review. Also, the audits were conducted on random weekdays and once on Sundays. In order to evaluate road user safety without bias based on the type of vehicle, audits were done on a two-wheeler without gear and on a bicycle. The checklist was used to record the observations, and issues were labeled as Yes, No, or NA. A score was assigned to checklist observation to convert qualitative data to quantitative data [15]; a score of “1” was assigned for “Yes,” and a score of “0” for “No.” If the answer to a direct question, such as “Is the pedestrian facility available?” is yes, the score is “1,” and if the answer to an indirect question, such as “Is road having dangerous pot-holes?” is yes, the score is “0,” as it can cause an accident, so a “0” score is given for the element that can cause an accident. If a question does not apply to a section, it is marked as “NA,” and no score is assigned to it; for example, if a flyover is not present, it is marked as “NA”. If there are 5 questions in the head “Visibility and Intersection”, and 2 of them have scored “1”, other 2 have scored “0,” and 1 is “NA”, then to convert to the score of “10”:

Identification of Infrastructural Causative Factors for Road Accidents …

=

547

Sum of all the element with score 1 x10 Total no.of element − No. of an element with"NA"

Hence, the head “Visibility and intersection” score will be 5 out of 10. The higher the score, the better will be the road section in terms of safety and the lesser probability of accidents happening on that road.

4.4 Questionnaire The questionnaire survey was conducted to validate the result, which will be obtained from the road safety audit. The questionnaire is totally based on the road safety audit checklist questions. In the questionnaire, the Likert scale with five responses was used to quantitatively get the likelihood of the accident different elements of the road can cause. 0 being accident less likely and 5 being accident more likely. The sampling method that is used is the stratified sampling method. The probable respondent should have worked in areas of transport engineering and preferably worked in the area of road safety or academicians who teach transportation engineering, or the traffic police, who are involved in accident investigations, but to eliminate the biases, the police are eliminated as we have seen in the literature review that they point towards the road user as the primary cause of the accident. So, for the questionnaire, the respondents selected are academicians or transport engineers working in the industry (professionals). The number of respondents was chosen according to the Table 2. Set 1 is selected for the questionnaire survey; 20 responses will be required. After the determination of the sample size of each stratum, further survey will be done Table 2 Determination of sample size Head

Formula

Symbol

Variable

Set-1

Determination of sample size ‘n’

n= [16]

z

Confidence level (Value obtained from Z-value table)

1.96

e

Margin of error

5%

p

Unknown population 0.5 proportion (Taken so that the value of P(1-P) becomes maximum)

n

Desired sample size

Stratified sampling

p(1-p)(z/e)2

384.1

5% of n

N

~ 20

Academician (50%)

S1

10

Professional (50%)

S2

10

548

P. Pasawala and B. Shah

with the help of judgmental sampling. The questionnaire was floated to the academicians with master/Ph.D. in transportation engineering and the professionals with any experience in the industry having master/Ph.D. in transportation engineering. The name and the institute where the respondent works are kept confidential to maintain the anonymity of the respondents. Based on the profile of the respondents, 40% of the professionals were of 0– 5 years of experience, which primarily includes working professionals, 20% of the respondents are of 5–10 years of experience, 25% of the respondents are of 10– 15 years of experience and remaining 15% of the respondents are of 20–25 years of the experience. The questionnaire responses were recorded on a scale of 5; for analysis, the score of 5 was converted to a score of 10 using a weighted average. E5 Formula : W =

Wi x Ni E ; N

i=1

wnere, Ni = number of responses for Wi ; N = Total responses; = Wi Assigned score. (From 1 to 5); W = Weighted average.

4.5 Crash Data After the road safety audit for the selected stretch was completed, secondary data on crashes in the previous three years (2018–2020) was collected from the police station, and the accident spots were marked on the map using the longitude and latitude registered in the police data. Then the number of accidents was counted for the 500 m road section for which the road safety audit was conducted. The accident on the selected road stretch for the past 3 years for the different sections of 500 m is (Table 3).

4.6 Data Analysis For the results, the value of Pearson’s correlation coefficient R can be between –1 and 1, in which + 1 signifies the positive correlation and –1 signifies the negative correlation. For the analysis of the collected data and to establish the correlation, the results were obtained from a scale of 10, in which the score of 0 means the section of the selected road is more prone to accidents and 10 means the road is less prone of the accident, for the analysis purpose the scale is reversed, which means 0 means the section will be less prone to the accident and the 10 will be more prone to accident. The following step is done to avoid confusion as when we do correlation on the same

Identification of Infrastructural Causative Factors for Road Accidents …

549

Table 3 No. of accident on the selected road Section

No. of Accident

Section

No. of Accident

Section 1

2

Section 18

7

Section 2

3

Section 19

6

Section 3

2

Section 20

7

Section 4

3

Section 21

5

Section 5

3

Section 22

3

Section 6

1

Section 23

1

Section 7

2

Section 24

1

Section 8

1

Section 25

0

Section 9

4

Section 26

6

Section 10

3

Section 27

1

Section 11

2

Section 28

1

Section 12

3

Section 29

2

Section 13

1

Section 30

0

Section 14

3

Section 31

0

Section 15

3

Section 32

0

Section 16

4

Section 33

0

Section 17

6

observation without inverting the scale, and if we get -1 as the correlation coefficient will mean that with the decrease in the section’s score, there will be an increase in the accidents, and to avoid such confusion the scale was reversed. The weights were assigned to every observation of the road sections: Weight for the safety audits :

1 1 = = 0.125 N o. o f heads 8

Weight for the expert opinion :

Scor e o f the 1 head Sum o f scor e o f all head

Here, Head = group of similar elements of the road, for example, visibility and intersection, speed limit, and others, as mentioned in Sect. 4.2 and (Table 4). Hence, the final score of all the heads for all the sections = score of the head x Weight of the head. E (xi − x)(yi − y) The formula for correlation : r = √ E (xi − x)2 (yi − y)2 where, r = correlation coefficient.

550

P. Pasawala and B. Shah

Table 4 Weight for calculating the weighted score Issue

Weight from Audit

Weight from a Questionnaire survey

Visibility and Intersection

0.12

0.13

Pedestrian and cyclist

0.12

0.12

Speed limit

0.12

0.12

Sign, pavement marking, and delineation

0.12

0.11

Cross-section and alignment

0.12

0.11

Road-side hazard

0.12

0.14

Light and Night-time issue

0.12

0.13

Miscellaneous

0.12

0.13

Table 5 Correlation matrix–Road safety audit & accidents

Heads (Refer 4.2) 1 2 3 4 5 6 7 8 Accident

1

2

1.0 0.5 0.2 0.2 0.1 0.0 -0.1 0.0 0.5

1.0 0.3 0.5 -0.2 0.3 0.1 0.2 0.4

3

1.0 0.2 0.1 0.2 0.1 0.1 0.3

4

5

6

1.0 -0.3 0.2 0.2 0.4 0.0

1.0 -0.3 -0.2 -0.2 0.3

1.0 0.4 0.5 0.00

7

1.0 0.3 0.0

8

1.0 0.0

Accident

1.0

xi & yi = Value of variable X & Y; for example, if correlation must be found between Accident and Visibility and Intersection head, then the X & Y will be the weighted score of the respective variable; x, y = Mean of the value of X & Y variable (Tables 5 and 6). Now, with the help of the strongly and moderately correlated factors, prioritization of the stretches E was done with the help of the Relative Importance Index: Formula: (score of the head questionnaire x Weight of the head) for the strongly or moderately related element of the road (Table 7).

5 Results and Discussions In the data analysis section, the data collected from the road safety audit which is conducted by the researcher and the data from the questionnaire survey, in which the respondent was experts in the field of transport engineering was done (Refer Tables 6

Identification of Infrastructural Causative Factors for Road Accidents …

551

Table 6 Correlation matrix–A questionnaire survey & accidents Heads (Refer 4.2)

1

2

3

4

5

6

7

8

Accident

1

1.0

2

0.5

1.0

3

0.2

0.3

1.0

4

0.2

0.5

0.2

1.0

5

0.1

–0.2

0.1

–0.3

1.0

6

0.0

0.3

0.2

0.2

–0.3

1.0

7

–0.1

0.1

0.1

0.2

–0.2

0.4

1.0

8

0.0

0.2

0.1

0.4

–0.2

0.5

0.3

1.0

Accident

0.5

0.4

0.3

0.0

0.3

0.00

0.0

0.0

1.0

Table 7 Prioritization of the road section from the co-related element of the road Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Section Section-15 Section-18 Section-9 Section-19 Section-17 Section-14 Section-7 Section-2 Section-3 Section-16 Section-20 Section-1 Section-13 Section-6 Section-12 Section-32 Section-4

Priority to be given 3.71 3.64 3.39 2.88 2.86 2.81 2.72 2.70 2.50 2.42 2.42 2.42 2.34 2.29 2.24 2.15 2.12

Rank 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Section Section-24 Section-29 Section-26 Section-22 Section-23 Section-5 Section-11 Section-10 Section-28 Section-31 Section-21 Section-25 Section-8 Section-30 Section-27 Section-33

Priority to be given 2.09 2.08 1.89 1.83 1.83 1.82 1.82 1.67 1.53 1.38 1.33 1.21 0.93 0.74 0.55 0.14

and 5); the results show that both the results are highly correlated and hence we can say the research has high validity. From the data analysis Table 5, it can be clearly seen that there is a positive correlation between the different elements of the road, which are related to visibility and intersection; Pedestrian and cyclist; Speed limit, sign pavement marking, and delineation; cross-section, and alignment; Miscellaneous, and Light and night-time hazards are positively correlated with the accidents, whereas road-side hazard shows no relation with the accidents. There is a positive correlation between the various types of road elements and accidents, but only two are strongly correlated: pedestrian and cyclist and sight distance and intersection. Elements related to speed limit and cross-section and alignment

552

P. Pasawala and B. Shah

are moderately related; otherwise, no strong relationships can be seen in the other elements, which means that if we focus on the elements related to pedestrians and cyclists; sight distance and intersection; cross-section and alignment; speed limit, there will be a drastic decrease in road accidents caused by road elements, increasing the safety of road users. From RSA, we can observe that the main factors in the Visibility and Intersection are Inadequate sight distance, U-turn provision on an unsignalized intersection, and unsignalized intersection. The main factor in the pedestrian and cyclist head are not providing facilities like footpaths and cycling tracks and not providing enough crossing points in the midblock. The main factor in speed limit is not providing speed calming measures, and in cross-section and alignment, the main factor is not having a sign before enough distance of flyover so that road users can decide on which road to go. From Table 7, we can see that Section 15, Section 18, and Section 9 are the first 3 roads to be prioritized, so if we apply the corrective measures on the prioritized road section, then there is the possibility that the accident on that section will be decreased. Hence, there will be an overall decrease in road accidents as these are the sections where most of the accidents occurred in the past. It can also aid the relevant urban authority (here in this case–Municipal Corporation) in prioritizing the road sections for implementation of road safety engineering interventions. A sign showing a flyover ahead shall be provided before an adequate distance from the flyover so that the driver can make an informed decision beforehand. Speedcalming measure like speed humps has to be provided, along with the proper delimitation and adequate crossing in mid-blocks and near BRTS stands. All the sight distance blocking elements like hoarding, illegal structures shall be removed from the road. Special cycling lanes and pedestrian lanes shall be provided alongside the road. The intersection with heavy traffic flow shall be signalised. These are the measures to make the roads forgiving and reduce the accidents, but there should also be strict enforcement of the laws on human factors such as overspeeding, using the wrong lane, driving vehicles without proper safety equipment like helmets and seat belts, etc.

References 1. Global Status Report on Road Safety (2018) World Health Organization, Geneva 2. Zou X, Yue WL (2017) A bayesian network approach to causation analysis of road accident using Netica. J Adv Transp 2017:1–18 3. World Health Organization webpage. https://www.who.int/news-room/fact-sheets/detail/roadtraffic-injuries. Last accessed 20 Jun 2022 4. Mohan D, Tiwari G, Bhalla K (2020) Road safety in India report, Transport research and injury prevention programme. Indian Institute of Delhi, Delhi 5. Road accident in India. Ministry of Road Transport and Highway, Transport research wing, Delhi 6. Ahmedabad urban road accident study (2016) JP Research India Pvt. Ltd., Ahmedabad

Identification of Infrastructural Causative Factors for Road Accidents …

553

7. Islam MH, Teik Hua L, Hamid H, Azarkerdar A (2019) Relationship of accident rate and road geometric design. In: Suistanable and Constrution Engineering Conference, pp 1–11 8. Chen H, Zhao Y, Ma X (2020) Crititcal factor analysis of severe traffic accident based on bayesian network in China. J Adv Transp:1–15 9. Jain SS, Singh PK, Parida M (2011) Road safety audit for four lane national Highway. Int Road Conf Road Saf Simul:1–22 10. Shenker R, Chowksey A, Sandhu HA (2015) Analysis of relationship between road safety and road desing parameter of four lane national highway in India. IOSR J Bus Manag:60–70 11. Roudini S, Keymanesha M, Ahangar AN (2017) Identification of “Blackspot” without using accident information. Bull Soc R Sci Liege:667–676 12. Desai MM, Patel AK (2011) Road Accident study based on regression model: A case study of Ahmedabad. Natl J Recent Trends Eng Technol 13. Swamy S, Bhanuki N, Sinha S (2019) Managing road safety in Ahmedabad. Transport and communication bulltetin for Asia and Pacific 14. MoHUA; UNDP (2013) Development of toolkit under Suistainable urban transport project. Ministry of Housing and Urban Affair, Delhi 15. Jones JR (2013) A method to quantify road safety audit data and results. UTAH STATE UNIVERSITY, Logan, Utah 16. Hoggs R, Tanis E, Zimmerman D Probability and statistical inference, 9th edn

Modelling Longitudinal and Lateral Vehicle Movement Behavior Under Multiple Influencing Vehicles Dhiraj Kinkar, Madhu Errampalli, Mukti Advani, and Saraswathi Sethi

Abstract In a microscopic traffic analysis, realistic modelling of vehicular movements in mixed traffic conditions is most important to arrive at outputs with adequate accuracy. However, the majority of the studies considered single leader vehicles as influencing vehicles to estimate the following vehicle movement which may not be a realistic behaviour under heterogeneous and non-lane discipline traffic circumstances. Moreover, both longitudinal and lateral movements take place simultaneously in such non-lane discipline situations. Considering these, the present study emphasizes the significance of the influence of other surrounding vehicles on the subject vehicle while modelling its movement and developing longitudinal (carfollowing) and lateral (lane change) movement models of a motorized vehicle with heterogeneous traffic conditions. For this, vehicular trajectory data was extracted from videography data from nine mid-block road sections. Accordingly, relevant independent variables of surrounding vehicles along with subject vehicle characteristics and road geometry were considered to develop car-following and lane change models for estimation of longitudinal and lateral movements respectively. The developed models have been validated using the observed data which can be utilized for estimating longitudinal and lateral vehicular movements under prevailing mixed traffic conditions with reasonable accuracy. Keywords Car following · Lane changing · Heterogeneous traffic · Multiple influencing vehicle

D. Kinkar · S. Sethi National Institute of Technology, Kurukshetra, Haryana, India e-mail: [email protected] M. Errampalli (B) · M. Advani CSIR-Central Road Research Institute (CRRI), New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_44

555

556

D. Kinkar et al.

1 Introduction 1.1 General In India, due to mixed traffic conditions and poor traffic discipline, traffic management becomes a big challenge for the authorities and planners in providing efficient vehicular traffic operations. The presence of prevailing mixed patterns of traffic, lane indiscipline and also high traffic volume on roads needs a good consideration of data and its interpretation along with appropriate traffic flow analysis. Since vehicles of various types are allowed to mix and share the same road space along the length of a carriageway, traffic analysis becomes more complex compared to homogeneous traffic conditions. Even traditional field data in such situations are generally not suitable for studying vehicle interactions with variations of traffic volume and vehicle composition on stream speed and capacity. Researchers have studied alternate methods and techniques for measuring transportation network performance considering various parameters including travel speed, delay, level of service, connectivity, safety, link performances, volume-capacity ratio, etc. Since the past few decades, traffic simulation technique has become popular among traffic and transportation professionals, a large number of microscopic traffic simulation models have been developed worldwide, but very few focus on typical mixed traffic comprising the space sharing based traffic behavior. In view of this, there is a high need for an appropriate microscopic traffic simulation model that can estimate driver behavior precisely and realistically focusing on mixed traffic conditions. While developing a microscopic traffic simulation model, core inbuilt traffic models like the car-following model and lane-changing model for the urban road network need to consider these mixed traffic conditions and also able to evaluate a wide range of transport policies. However, the majority of the studies considered single leader vehicles as influencing vehicles to estimate the following vehicle movement which may not be able to estimate realistic behaviour under heterogeneous and non-lane discipline traffic circumstances. Moreover, both longitudinal and lateral movements take place simultaneously in such non-lane discipline situations. Considering these, the present study proposes to develop longitudinal (carfollowing) and lateral (lane change) movement models of motorized vehicles with heterogeneous traffic conditions under the influence of other surrounding vehicles on the subject vehicle. Accordingly, relevant independent variables of surrounding vehicles along with subject vehicle characteristics and road geometry have been considered to develop car-following and lane change models for estimation of longitudinal and lateral movements respectively. The developed models were further validated using the observed data in order to utilize for estimating longitudinal and lateral vehicular movements under prevailing mixed traffic conditions with better accuracy.

Modelling Longitudinal and Lateral Vehicle Movement Behavior Under …

557

1.2 Longitudinal and Lateral Movements Gipps [1] developed a framework to model lane-change decisions in urban driving conditions, however, presupposes that a lane change manoeuver will only happen when it is safe, that is, when there is sufficient space in the target lane from the leader vehicle. Under heterogeneous and non-lane discipline traffic circumstances, estimation of vehicular movements depends not on a single leader vehicle, but surrounding vehicles that are travelling close to it. So, not only the longitudinal gap, but the vehicle tries to adjust lateral gaps with these surrounding vehicles in order to maintain safety while moving on roads. This would be different for motorized two-wheelers (MTW) as the width of the vehicle is small and tries to squeeze in between adjacent and leading vehicles. Mallikarjuna et al. [3] carried out an analysis of the lateral dispersion of vehicles on 10-m-wide highways and studied the effect of the lateral positions of the leading and following vehicles on the longitudinal space. Relationships between the lateral space and area occupancy are displayed for various vehicle combinations. In addition to vehicle type and speed, it has been discovered that staggering affects the longitudinal spacing that cars maintain. Malikarjuna et al. [2] examined lateral gap-keeping behavior in mixed traffic conditions and found that the lateral distance between the subject vehicle and the adjacent vehicle increases as the speed of the adjacent vehicle increases. Kanagaraj et al. [4] studied the characteristics and factors relating to the longitudinal and lateral movement of the vehicles and demonstrated that MTW moves also in the lateral direction. Asaithambi and Basheer [5] studied car-following behavior in mixed traffic situations under different vehicle classifications and different follower-leader pair types to investigate the vehicle-following behavior. The analysis demonstrates that the size of the leader has a greater influence on the follower’s decision to stagger; when the leader’s speed is higher, the following vehicle has more freedom to choose a different manoeuver; heavy vehicles as the pursuers are unwilling to change the manoeuver in the shorter time interval. A multinomial linear regressive model was created to account for vehicle following behavior. The macroscopic and microscopic features of mixed traffic on MTW and cars were studied by Wong et al. [7] to determine all obtrusive vehicle movements and derive specific features for various vehicle modes, lanes, and density levels such as lateral positions, lateral gaping, and longitudinal gaping. Raju et al. [6] studied driver behavior in mixed traffic on multi-lane expressways to conclude that smaller vehicles switch leader vehicles more frequently to avoid delays, which leads to shorter following and perception times and more aggressive gap acceptance. From there, it has been observed that the datasets of most studies are limited to one or two locations and models were developed for heterogeneous traffic based on single leader vehicles; however, the effect of other surrounding vehicles which is also influential in heterogeneous traffic with no-lane discipline behavior (i.e., space sharing based behavior) is not considered. This highlights the need to consider the different vehicles and their types in the surroundings of the subject vehicle.

558

D. Kinkar et al.

1.3 Objectives, Scope and Methodology The main objectives of the present study are: • Analyzing the longitudinal and lateral behavior of drivers considering surrounding vehicles on the mid-block sections in mixed traffic conditions. • Development of a car following model for mixed traffic conditions under the conditions of having multiple influencing vehicles and no influencing vehicle (i.e., free flow) of motorized vehicles at the mid-block section. • Development of lane changing model for mixed traffic conditions of motorized vehicles at mid-block section. The scope of this study covers mid-block road sections. For this, detailed data collection, extraction, and analysis have been carried out to estimate vehicular movement in response to surrounding influencing vehicles. Further, the scope of this study is limited to motorized vehicles only.

2 Methodology and Data The present study intends to develop various models for vehicular movement in urban mixed traffic conditions. Therefore, a detailed analysis of vehicle trajectory data of traffic flow has been carried out. The data is mainly extracted for the subject vehicle and influencing vehicles. These influencing vehicle(s) can be on any side of the subject vehicle and the proximity area around the subject vehicle is divided into four zones as presented in Fig. 1. Fig. 1 Depiction of the 4 zones adopted for analysis around the subject vehicle

Zone 2

Subject Vehicle Zone 3

Zone 1

Zone 4

Modelling Longitudinal and Lateral Vehicle Movement Behavior Under …

559

Keeping the objectives in focus, detailed data collection and extraction methodology have been proposed. As mentioned in the scope of this study, only mid-block road sections have been considered and a total of nine mid-block locations in four different cities of Ahmedabad, Meerut, Delhi and Hyderabad have been considered with varying road widths as given in Table 1. To obtain the data, a video camera was mounted at a vantage point to collect the vehicular movement behaviour. After the collection of traffic data, extraction of the collected data using the traffic data extractor (TDE) which is a semi-manual software package developed by IIT Bombay (TDE User Manual). The recorded videos have been imported into the TDE tool and calibrated using the marked rectangle. The road stretch is described by specifying the entry line and the exit line. The length and width of the stretch are entered as sides of a rectangle. Adopting this methodology, the trajectory data have been extracted using the TDE software tool. The sample size for different vehicle types is shown in Table 2. Since the modelling of vehicular movements is the intent of the study, the vehicle position in each frame is considered as one sample. A total of 49,251 frame counts have been extracted for different vehicle types which have been further analysed to understand vehicle behaviour and modelling vehicular movements. From these data, the significance of zones and the different influencing variables of each of these zones surrounding the subject vehicle have been analyzed. Table 1 List of mid-block locations considered in the study

S. no

Location name

Road width

1

Harkesh Nagar, Mathura 11.2 m Road, New Delhi

2

Rajaram Kohli Marg, New Delhi

7.5 m

3

Balaji Temple Bypass Road, Hyderabad

9.0 m

4

Vijay Char Rasta, Ahmedabad

10.5 m

5

Paldi Road, Ahmedabad 8.0 m

6

Narol Sarkej Road, Ahmedabad

17.0 m

7

Garh Road, Meerut

8.0 m

8

Swastik Road, Ahmedabad

7.0 m

9

Maharani Bagh Road, New Delhi

10.0 m

560

D. Kinkar et al.

Table 2 Extracted sample size of motorized vehicles on mid-block locations Vehicle type

Sample size (Frame count)

Auto Rickshaw

5803

Bus

728

Car

16,402

Battery Cycle Rickshaw (CRB)

876

Heavy Commercial Vehicle (HCV)

799

Light Commercial Vehicle (LCV)

2275

MTW

21,921

Multi-Axle Vehicle (MAV)

447

Total

49,251

3 Data Analysis

10 0 Left Centre Right Lateral Position on Carriageway

24.86

68.27 24.48 23.56

Auto 54.27 48.19

MTW

21.25 28.25

20

80 70 60 50 40 30 20 10 0

6.87

31.49

44.2 43.12

Car

Percentage (%)

30

Auto 38.68 35.22

40

MTW

24.35

Car 50

25.44 31.45 35.64

Average Speed (km/hr)

Analysis has been carried out to find out speed variation with respect to the lateral displacement of the vehicle. Since there is no specific lane discipline, the road width has been laterally divided into three equal areas (left, centre and right) in order to understand vehicular behaviour with respect to placement on the road. Left area is near to shoulder/footpath and right area is near the median side. Figure 2 presents the average speed characteristics of Car, MTW and Auto moving in these left, central and right areas on the road. As can be seen in Fig. 2, the speed of Car, MTW and Auto moving in the right area, i.e. near the median side is higher compared to the other two areas, i.e. Left and Centre. Further analysis has been done on lateral placement distribution as shown in Fig. 2. It can be seen from Fig. 2 that Cars have a high tendency to travel on right rightmost side of the road whereas MTW and Auto have a high tendency to travel on the central part of the road.

Left Centre Right Lateral Position on Carriageway

Fig. 2 Average speed and percentage distribution with respect to lateral placement on the carriageway

Modelling Longitudinal and Lateral Vehicle Movement Behavior Under …

561

4 Car-Following Model 4.1 Subject Vehicle with Surrounding Vehicles The present study intends to develop car following models to estimate vehicular longitudinal movement for urban traffic conditions within mixed traffic. Therefore, a detailed analysis of vehicle trajectory data of traffic has been carried out to prepare the data sets. The data was mainly extracted for two types of vehicles, viz. subject vehicles and influencing vehicles. Influencing vehicle(s) can be in any of the four zones surrounding the subject vehicle as shown in Fig. 1. The car-following model for motorized vehicles as subject vehicles with all surrounding influence vehicles has been developed considering various independent variables utilising multiple linear regression modelling. The independent variables considered are: Speed of the subject vehicle in the previous time interval, speed of the influence vehicle in that zone in the previous time interval, Lateral distance from the median for the subject vehicle in the previous time interval, Passenger car units of subject and influence vehicle to consider mixed traffic in terms of vehicle type, Road width and gap between subject and influence vehicle. The format of the proposed car-following model is given below. ViS (t) = a1 ∗ ViS (t − 1) + a2 ∗ DMiS (t − 1) + a3 ∗ PCUS + a4 ∗ RW + a5 ∗ PCUiIV + a6 ∗ ViIV (t − 1) + a7 ∗ RDi

(1)

where V is speed in m/s Subscripts ‘S’ and ‘IV ’ represent the subject vehicle and the influencing vehicle Superscript ‘i’ represents zone (1-left; 2-front; 3-right; 4-back) ‘t’ and ‘t–1’ are current and previous time intervals, respectively DM is the lateral distance from the median for the subject vehicle RW is Road width in m PCU is Passenger car unit RD is the relative distance (gap) between the subject and the influencing vehicle in m a1 , a2 , a3 , a4 , a5 , a6 , and a7 are regression coefficients to be estimated. The car-following equations for each zone are developed separately as given below: VS1 (t) = 0.872 ∗ VS1 (t − 1) + 0.018 ∗ DMS1 (t − 1) − 0.099 ∗ PCUS + 0.020 ∗ RW + 0.047 ∗ PCUiv1 + 0.080 ∗ Viv1 (t − 1) + 0.005 ∗ RD1 · · · (R2 = 0.97)

(2)

562

D. Kinkar et al.

VS2 (t) = 0.882 ∗ VS2 (t − 1) + 0.029 ∗ DMS2 (t − 1) − 0.099 ∗ PCUS + 0.053 ∗ RW + 0.032 ∗ PCUiv2 + 0.070 ∗ Viv2 (t − 1) + 0.013 ∗ RD2 · · · (R2 = 0.98)

(3)

VS3 (t) = 0.873 ∗ VS3 (t − 1) − 0.003 ∗ DMS3 (t − 1) − 0.116 ∗ PCUS + 0.033 ∗ RW + 0.007 ∗ PCUiv3 + 0.093 ∗ Viv3 (t − 1) + 0.086 ∗ RD3 · · · (R2 = 0.97)

(4)

VS4 (t) = 0.891 ∗ VS4 (t − 1) − 0.012 ∗ DMS4 (t − 1) − 0.062 ∗ PCUS + 0.040 ∗ RW + 0.013 ∗ PCUiv4 + 0.062 ∗ Viv4 (t − 1) + 0.031 ∗ RD4 · · · (R2 = 0.98)

(5)

The final resultant speed for the subject vehicle is considered as the average of speeds estimated in each of the zones as given below: 1  i ∗ V (t) n i=0 S n

VS (t) =

(6)

where n–Number of zones (left, front, right and back) The R2 values of Eqs. (2)–(5) show good statistical validity, hence it is considered that the developed car-following equations would be able to explain the vehicular movements with reasonable accuracy.

4.2 Subject Vehicle with No Surrounding Vehicles Many times, the subject vehicle will be travelling on a link without having any influence from other vehicles because there will not be any vehicle present in the surrounding influence area. The car-following model for motorized vehicles as the subject vehicle with no influence vehicles in the influence area has been developed considering various independent variables utilising multiple linear regression modelling. The independent variables considered are: Speed of subject vehicle in the previous time interval, Maximum acceleration of subject vehicle, Lateral distance from median for subject vehicle in the previous time interval, Passenger car unit of subject vehicle, Road width and Free speed of subject vehicle. The developed car-following equations are given below:  VS (t) = VS (t − 1) +

0.048 ∗ aS + 0.019 ∗ DMS (t − 1) + 0.001 ∗ PCUs + 0.048 ∗ RW



Modelling Longitudinal and Lateral Vehicle Movement Behavior Under …

     VS (t − 1) VS (t − 1) 2.975  2 ∗ 1− R = 0.97 ∗ 1.15 + f f vS vS

563

(7)

wherev f is free speed in m/s a is maximum acceleration in m/s2

4.3 Validation In order to validate the above-developed equations, the predicted speed from Eq. (6) is compared with the observed speed in the field. A comparison graph has been plotted between the observed speed and predicted speed for the validation of car-following model with an influence vehicle as shown in Fig. 3a. Root Mean Square Error (RMSE) value for this model is found to be 1.45 m/s. From this, it can be concluded that the developed car-following model with surrounding influencing vehicles is able to predict the speed of the subject vehicle realistically with adequate accuracy. Similarly, in order to validate the developed car-following model with no influence vehicle, the predicted speed from Eq. (7) is compared with the observed speed in the field. A comparison graph has been plotted between the observed speed and predicted speed for the validation of the car following the model with no influence vehicle as shown in Fig. 3b. The RMSE value for this model is found to be 0.83 m/s. From this, it can be concluded that the developed car-following model with no influencing vehicles (free flow) is able to predict the speed of the subject vehicle realistically with adequate accuracy.

30 Preedicted Speed (m/s)

Preedicted Speed (m/s)

30 25 20 15 10 5 0

25 20 15

10 5 0

0

5 10 15 20 25 Observed Speed (m/s) (a) With surrounding vehicles

30

0

5 10 15 20 25 Observed Speed (m/s)

(b) With no surrounding vehicle

Fig. 3 Observed and predicted speeds of subject vehicle in different conditions

30

564

D. Kinkar et al.

5 Lane Change Model

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

Lateral Shift (m)

Lateral Shift (m)

From the data, the vehicle lateral movement behaviour in terms of lateral shift (difference in lateral position in two successive time intervals is measured in m) has been analysed with respect to the speed of the subject vehicle. The variation of lateral shift for different speeds of Car is shown in Fig. 4. From Fig. 4, it can be observed that the lateral shift of cars is more at slower speeds (70 km/hr). As the speed of the car increases, lateral shift decreases up to 0.15 m at 25 km/hr and increases up to 0.4 m at 95 km/hr. It can also be seen from Fig. 4, as the speed of MTW increases, lateral shift increases up to 0.4 m at 95 km/hr. As the speed of the auto increases, lateral shift increases up to about 0.18 m at 40 km/ hr and reduces subsequently as shown in Fig. 4. The developed lane change equations to estimate lateral shift for different vehicles are given below:

20 40 60 80 Car Speed (km/hr)

Lateral Shift (m)

0

100

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

20 40 60 80 MTW Speed (km/hr)

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

20 40 60 80 Auto Speed (km/hr)

100

Fig. 4 Relationship between speed and lateral position of different vehicles

100

Modelling Longitudinal and Lateral Vehicle Movement Behavior Under …

565

Car XLat (t) = 0.2489 + 0.00008 ∗ VCar (t − 1)2   − 0.0058 ∗ V(t − 1) R2 = 0.96

(8)

MTW XLat (t) = 0.197 + 0.00004 ∗ VMTW (t − 1)2   − 0.0015 ∗ VMTW (t − 1) R2 = 0.99

(9)

Auto XLat (t) = 96.384 − 0.0485 ∗ VAuto (t − 1)2   + 3.7976 ∗ VAuto (t − 1) R2 = 0.87

(10)

where ΔX Lat is the lateral shift in m.

6 Conclusions In this study, the longitudinal and lateral behaviour of vehicles under multiple influencing vehicles have been analysed and car-following and lane change models have been developed. The summary of findings from this study has been given below: • Lateral placement distribution analysis found that Cars have a high tendency to travel on the right side of the road whereas MTW and Auto have a high tendency to travel on the central area of the road. • Speed analysis found that the speed of Car, MTW and Auto moving on the right side of the road is higher compared to the other two areas, i.e. Left and Centre. • Car following model has been developed under mixed traffic conditions considering the surrounding vehicles and no surrounding vehicle case for mid-block location. • Vehicle behaviour in terms of lateral shift has been analysed with respect to speed and found that lateral shift in case of car is more in slower speeds (70 km/hr). As the speed increases, lateral shift decreases up to 0.15 m at 25 km/hr and increases up to 0.4 m at 95 km/hr. the speed of MTW increases, lateral shift increases up to 0.4 m at 95 km/hr. As the speed of the Auto increases, lateral shift increases up to about 0.18 m at 40 km/hr and reduces subsequently. • Lane change model has been developed to estimate lateral shift under mixed traffic conditions for mid-block locations. The developed car-following and lane change models in this study would eventually improve the predictions thus realistic evaluation of transport policies through microscopic simulation. In the future, it is also proposed to consider other vehicle types in modelling longitudinal and lateral behaviour so that predictions can become more realistic.

566

D. Kinkar et al.

References 1. Gipps PG (1986) A model for the structure of lane-changing decisions. Transp Res Part B: Methodol 20(5):403–414 2. Mallikarjuna C, Tharun B, Pal D (2013) Analysis of the lateral gap maintaining behavior of vehicles in heterogeneous traffic stream. Procedia–Soc Behav Sci 104:370–379 3. Mallikarjuna C, Kalaga RR, Seethepalli NVSK (2010) Analysis of microscopic data under heterogeneous traffic conditions. Transport 25(3):262–268 4. Kanagaraj V, Asaithambi G, Toledo T, Lee T-C (2015) Trajectory data and flow characteristics of mixed traffic. Transp Res Rec, J Transp Res Board 2491(1):1–11 5. Asaithambi G, Basheer S (2017) Analysis and modeling of vehicle following behavior in mixed traffic conditions. Transp Res Procedia 25:5094–5103 6. Raju N, Kumar P, Jain A, Arkatkar SS, Joshi G (2018) Application of trajectory data for investigating the vehicle behavior in mixed traffic environment. Transp Res Rec, J Transp Res Board 2672(43):122–133 7. Wong KI, Lee TC, Chen Y (2016) Traffic characteristics of mixed traffic flow in urban arterials. Asian Transport Studies 4(2):379–391

Comprehensive Analysis of Road Accidents and Surrogate Measures to Enhance Road Safety B. S. Jisha and M. Satyakumar

Abstract Worldwide, more than 50 million casualties occur in road crashes each year, in which 80% of road crashes are in developing countries. The country ranks one in the number of road accident deaths and an increase of about 47% in road crashes is expected in the next 20 years. Kerala is one of the top five accident-prone states in the country at the present time. Road safety becomes more and more important every year as the annual growth rate of traffic is more than 10% in Kerala. Road crashes tend to result in personal injury, loss of life, or damage to property. The challenging factors in the traffic conditions existing in our road networks are the mixed traffic conditions and vulnerable road users. The state needs a comprehensive road safety plan, which in turn requires enormous data with respect to the accident and its severity over a period of time in order to reduce the accident scenario. Contributing factors of road accidents need to be documented in order to decide the most appropriate solutions. The objective of this study is to analyze the accident data over a period of time, identify the causative factors, and suggest appropriate solutions. A 60 km stretch of National Highway NH 66 passing through the Kollam district of the state was selected for this purpose. Data from the State Crime Records Bureau (SCRB), over a period of three years from 2017 to 2019 was collected and analyzed. Year-wise accident statistics, trends and causative factors were identified. Certain locations were identified to be blackspots as per MoRTH standards were later taken up for detailed study and appropriate solutions were suggested. Keywords Road safety · Road crashes · Crash severity

B. S. Jisha (B) · M. Satyakumar Department of Civil Engineering, Mar Baselios College of Engineering and Technology, Thiruvananthapuram, Kerala 695015, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_45

567

568

B. S. Jisha and M. Satyakumar

1 Introduction Road safety refers to the methods and measures used to prevent road users from being killed or injured in road accidents or road crashes. Traffic safety is a major concern in developing countries like India. The most common measures to define road safety are the number of road crashes, number of road casualties and other negative consequences of road crashes. India ranks 1 in the number of road accident deaths across 199 countries reported in the World Road Statistics, 2018. As per the World Health Organization, accident-related deaths are known to be the eighth leading cause of death and the first largest cause of death among children aged 5–14 and adults aged 15–29. Road accidents are the leading cause of injuries, death and disabilities. Hence, a comprehensive road safety plan is required in India to reduce the number of accidents and severity of accidents. Road accident tends to result in personal injury, loss of life or damage to property. Accident is an unpredictable event caused by one or a combination of multiple factors. These factors can be grouped into three main categories, namely infrastructure, road user behavior, environmental. While road accidents are unpredictable events, the intensity can be reduced to a certain extent by employing suitable road safety measures. Therefore, a systematic analysis of road accidents is essential. A systematic approach is required to identify the contributing factors so that the most appropriate treatments can be selected and implemented. It is necessary to come out with a more comprehensive and systematic road safety plan to effectively reduce the crash frequency and to ensure safer traffic on our road networks.

1.1 Literature Review Rolison et al. [8] The main causes of road accidents are identified by multiple sources of reviews: expert views of police officers, lay views of the driving public, and official road accident records. The results reveal potential underreporting of factors in existing accident records, identifying possible inadequacies in law enforcement practices for investigating driver distraction, drug and alcohol impairment, and uncorrected or defective eyesight. The need for the accident report forms to be continuously reviewed and updated to ensure that contributing factor lists reflect the full range of factors that contribute to road accidents is highlighted. The delay in completing accident report forms should be minimized, possibly by the use of mobile reporting devices at the accident scene. Hu and Xiang [7] from their study, it was concluded that the geometric characteristics of the study corridor were observed to be varied, which affects the operating speed thereby affecting the level of safety. The operating speed of the highway was found to be the most influential factor in all models developed. The relationship between contiguous elements was established by developing operating speed models for the curve and tangent section and was observed that the operating speed of one element

Comprehensive Analysis of Road Accidents and Surrogate Measures …

569

is associated with the other element. The developed crash prediction models gave insight that not only the segment under study have an influence on the crash but also succeeding and preceding sections have a significant role in the crash occurrence. Baktha [3], mobile apps have become an integral part of our daily lives due to the various functionalities that they offer. The developer should consider the challenges faced and try to overcome them by following the proper steps. Also, it is imperative for the developer to have an open mind and should be well apprised about the current technologies, requirements and events in the mobile application field. To build a successful, all the guidelines should be properly considered and followed appropriately to avoid the risk of losing users due to lamentable/falling apps which leads to a scope for future areas of research.

1.2 Objectives The objectives of the study are • • • •

To identify the hotspots from the accident records To find out the causative factors by detailed study To suggest surrogate measures to improve the safety in the selected stretch. To develop a mobile application to alert the drivers.

2 Methodology The methodology of the study is shown in Fig. 1.

3 Accident Scenario Total number of crashes occurred, total injuries and no. of deaths due to road crashes from the year 2009 to 2019 were collected from the State Crime Records Bureau. Figure 2 shows the plot of the total number of crashes, total injuries and the death rate from 2009 to 2019. The number of reported accident cases is increasing year by year in general. Even if the total number of accident cases shows variation, the death rate as well as total injuries due to accidents continuously increasing. It shows that the effectiveness of the safety measures adopted so far is questionable. Implementation of more effective safety measures is essential in the present scenario. Analysis shows that in most of the district’s number of crashes increases year by year. Our road network includes different types, geometrical and traffic characteristics of each type are different. In Kerala, the road network includes the National Highway, State Highway, Major District Roads, Other District Roads, Village Roads, and PMGSY Roads. Figure 2 shows the road network has only 1% of the road network

570

B. S. Jisha and M. Satyakumar

Fig. 1 Methodology

in Kerala is National Highways and 24% of the total road crashes occur in this category of road. In State Highways also, the crash frequency is more. National Highways and State Highways are the higher road categories, a greater number of road users and thereby a greater number of road accidents. The design speed and design specifications of both the categories are same. Figure 3 shows the road network and percentage distribution of crashes on different categories of roads. Traffic conditions existing in the road networks of Kerala are mixed in nature. The type of vehicle involved in the majority of accidents is two-wheelers. Two-wheeler users are the most vulnerable. 40% of the accidents occurred in the state are due to two-wheelers.

Comprehensive Analysis of Road Accidents and Surrogate Measures …

571

50000 40000 30000 20000 10000 0 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Total Injuries No. of accidents Death Fig. 2 Road crash scenario in Kerala from 2009 to 2019

58%

24% 18%

4% 2% 14% 15% 1%

NH SH District Roads Rural

NATIONAL HIGHWAYS STATE HIGHWAYS OTHER ROADS

64%

Urban Project Roads

Fig. 3 Category-wise road network and crash distribution

4 Study Stretch For detailed analysis, 60 km road was selected as the study stretch. The selected stretch is a portion of NH 66 passing through Kollam district. The traffic condition on the road is mixed in nature and it is a two-way two-lane undivided highway. The stretch includes straight portions, curves, bridges, intersections, urban roads, etc. The accident records of this stretch were collected from the State Crime Records Bureau. Three years’ accidents data was collected (Fig. 4).

572

B. S. Jisha and M. Satyakumar

Fig. 4 Study stretch

5 Characteristic Analysis Characteristic analysis of the accidents occurred was performed based on time of occurrence, category of vehicles, driver characteristics, climatic conditions. The severity of the accidents was also analyzed in the selected stretch.

5.1 Time of Occurrence Majority of accidents occurred between 3 and 9 pm. About 40% of crashes occurred during this time. Daytime 9 pm to 12 pm is also critical, about 20%. Traffic volume is higher during these hours and evening 6–9 pm the visibility of the drivers is restricted. However, 65% of total crashes occurred in the daytime, from 6 am to 6 pm, and 35% were during night, 6 pm–6 am. Even if the traffic volume is lesser after 9 pm 15% of accidents occurred from 9 pm to 6 am. If special attention should be given to making the drivers more vigilant during night driving this 15% can be reduced as the traffic volume is very less. Most of the accidents occur during the night because of fatigue or drowsiness (Fig. 5).

5.2 Category of Vehicles The traffic consists of different vehicle compositions. From the analysis, it is clear that the most vulnerable road users are two-wheeler drivers about 55% of vehicles

Comprehensive Analysis of Road Accidents and Surrogate Measures …

573

Fig. 5 Distribution of crashes during day and night

Night 35% Day 65%

Fig. 6 Percentage wise distribution based on category of vehicles

Day Night

Heavy vehicle 12% Four wheeler 26% Three wheeler 7%

Two wheeler Two wheeler 55%

Three wheeler Four wheeler Heavy vehicle

involved in accidents are two-wheelers and 26% are four-wheelers which includes cars, jeep and taxis (Fig. 6).

5.3 Driver Characteristics Driver characteristics, age of the driver and sex of the drivers of the vehicles are studied. Young drivers, age group of 18–29, make more accidents. This may be due to a lack of experience and immature behavior to traffic. Out of the drivers, 60% of drivers are male drivers on the selected stretch. Male drivers make more accidents compared to female drivers (Figs. 7 and 8).

5.4 Severity of Accidents The severity of accidents that occurred on the selected stretch is also studied. 74% of the accidents are grievous in nature (Fig. 9).

574

B. S. Jisha and M. Satyakumar

Fig. 7 Percentage of accidents based on age of the driver

10% 29%

61%

Young Drivers Middle aged Old

Fig. 8 Percentage distribution based on gender

18% Male

82% Female

Fig. 9 Severity of accidents

12% 14%

GREVIOUS

74%

DEATH MINOR

5.5 Climate The climatic conditions at the time of occurrence of the accidents were also studied. About 51% of the accidents occurred in clear sunny climates. From this, it is inferred that the worst climatic condition is not the main factor that leads to accidents. Drivers are more careful while driving the vehicles in worst climatic conditions. And also, the absence of two-wheelers during rainy times may also a reason behind this (Fig. 10).

Comprehensive Analysis of Road Accidents and Surrogate Measures … Fig. 10 Percentage wise distribution based on climatic conditions Very hot 5%

Other 14%

Sunny/ Clear 51%

Cloudy 13% Heavy rain 1% Light rain 10%Mist/ Fog 6%

575

Cloudy Heavy rain Light rain Mist/ Fog Sunny/ Clear Very hot Other

6 Detailed Investigation The critical locations (hotspots) were identified on these stretches in terms of crash frequency. A stretch of about 500 m in length in which either 5 road accidents (in all 3 years put together involving fatalities/grievous injuries) took place during the last three calendar years or 10 fatalities (in all 3 years put together) took place during the last 3 calendar years is noted as hotspot (MoRTH). 27 hotspots were identified in the selected stretch and listed in Table 1. These 27 hotspots comprise straight stretches, curved portions and intersections as listed above. Detailed investigation includes measuring the carriageway width, shoulder width, presence of road markings, sign-boards, signals, etc.

6.1 Straight Stretches 26 straight stretches and 14 intersections were analysed. Out of which, most of the stretch’s carriageway width is just sufficient but almost no shoulders or insufficient shoulder width. Out of the total locations investigated, about 70% of the test sections have no shoulders. Only paved or dense-graded shoulders were considered as shoulders. Grass and soil are not suitable driving surfaces and therefore normally do not function as shoulders. Traffic signage helps drivers remain aware of what’s coming up and what’s important on the road ahead. It’s one of the oldest ways of enforcing safety for both drivers and pedestrians on the road. In some portions absence of signboards, markings, speed breakers and other safety measures were missing.

576

B. S. Jisha and M. Satyakumar

Table 1 Identified hotspots in the selected stretch and its geometry

Sl:no 1 2 3 4 5 6 7 8 9 10 11

Place

13 14 15 16 17 18 19 20 21 22 23 24 25

CHANGANKULANGARA CHATHANNOOR CHAVARA CHINNAKKADA EDAPALLYKOTTA ITHIKARA KANNETTY KARUNAGAPALLY KMML JUNCTION KOTTANKULANGARA KOTTIYAM KSRTC JUNCTION (KNPLY) KUTTIVATTOM MEVARAM MYLAKKAD NEENDAKARA OCHIRA PALLIMUKKU PARAKKULAM PARIMANAM PARIPALLY POLAYATHODU PUTHENTHURA PUTHIYAKAVU THATTAMALA

26 27

UMAYANALLOR VAVVAKKAVU

12

Bridge

Straight ✔ ✔ ✔ ✔

Curve

✔ ✔ ✔ ✔

✔ ✔

✔ ✔ ✔ ✔ ✔ ✔

Intersection

✔ ✔

✔ ✔

✔

✔

✔

✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔

✔ ✔

✔

✔ ✔

✔ ✔ ✔ ✔ ✔ ✔ ✔

6.2 Curves Curved portions in the selected stretch were analysed. Sight distance, superelevation and speed on the curves are the most important factors associated with curves. Sight distance in most of the locations was sufficient. In horizontal curves accidents occur when the centrifugal force is more than the direction and momentum of the force, which makes the vehicle move in a straight line instead of curved path. Under estimation of speed may be a contributing factor while moving through curved roadways, which necessitates speed adjustment and thus causes fatal and serious injury accidents, especially in heavy vehicles. Approaching sharp curves without realising the speed and the driver fails to decelerate traffic incidents may occur. Warning signs are necessary to warn the driver that he is approaching the curves and has to move with utmost care. Warning signs were absent in most of the curves, shoulder width is also insufficient.

Comprehensive Analysis of Road Accidents and Surrogate Measures …

577

6.3 Bridges Three bridges are there on the list of hotspots. The carriageway width at these three locations is less than required and shoulder width is less for two bridges and no shoulder for one.

6.4 Traffic Characteristics Traffic Volume. During Traffic volume count was taken by videographic method. Peak hour traffic volume is also counted as peak hour, traffic volume is nearing its capacity. A spot speed study was also conducted. Operating Speed Speed limits indicate the maximum and minimum speed a driver can move in a certain category of road, assuming that there are favourable traffic and weather conditions. The 98th percentile speed was 65 KPH and the 85th percentile speed, limit for speed regulation was 50 KPH. Drivers may cause accidents by engaging in the following behaviours: Driving too fast, driving 5–10 kmph over the speed limit is not over speeding; however, even traveling just a few kmph over the speed limit can have serious consequences. Driving too slowly can also be as dangerous as driving too fast. If the vehicle moves 5–10 kmph or slower than surrounding traffic, you are more likely to be in an accident. Failing to maintain a reasonable speed; drivers who fail to consider weather conditions, traffic conditions, and roadway conditions while driving may end up in an accident, even if they are not technically exceeding the speed limit. From the detailed analysis it can be concluded that over speed, careless driving absence of warning signs are the main causes of accidents. The driver by knowing this hotspot region can reduce the speed of the vehicle and drive cautiously through the road and reducing the number of accidents. As smartphones become affordable for common people, they are starting to use them in all different parts of life. Nowadays, the use of smartphones with different mobile applications is common, especially in the transportation sector. These techniques combined with the right software can provide the user the location-based information, which can help in different ways.

7 SAAVADHAN -Mobile Application to Alert the Driver About Hotspots “SAAVADHAN” is a mobile application developed to alert the users about the hotspot regions. The app mainly focuses on the safety of the user using it. It alerts people about the accident-prone zones ahead of them using a beep sound. Thus, the drivers

578

B. S. Jisha and M. Satyakumar

Fig. 11 Logo of mobile app

slow down the vehicle and drive safely through the particular region. This helps in reducing the number of accidents occurring in that region. Driving safely and comfortably is one of the key factors every user needs or desires. “SAAVADHAN” provides safety through its efficient working and properties. The application indicates to the drivers about the accident-prone zones ahead of them when the vehicle reaches within 200 m of the area. Thus, by slowing down the vehicle the user could safely go through the particular region. It efficiently works for the safety of the users (Figs. 11 and 12).

7.1 Features of the App The application was developed in the expo platform. Expo is an open-source platform for making apps for Android, iOS, and the web with JavaScript. The main features of the mobile application are. Accurate GPS Location The application provides accurate GPS location. There is no delay in showing the respected hotspot region. The predicted location for a user is detected and spotted well by the app. It is a kind of helping tool for navigation. Alerts Only in the Direction of Driving The application provides specific alerts in the direction of the stretches chosen. The alert is in the form of a beep sound. Voice or Sound Alert When Approaching a Hotspot Region It alerts the driver by a beep sound around 200 m advanced of the hotspot. It would be safer for the users as it alerts them before reaching the accident-prone zone.

Comprehensive Analysis of Road Accidents and Surrogate Measures …

579

Fig. 12 Screenshot of the page showing hotspots

Safe Drive Safety during driving is one of the main features of the application. The beep sound alerts the drivers about the accident-prone zone which helps them drive safely through the respective region. It is also helpful during night hours when there is a possibility for the drivers to sleep while driving. The beeping sound helps them to be alert and thus avoid road accidents.

580

B. S. Jisha and M. Satyakumar

Offline Navigation The application not only provides online alerts but also works offline. Even if there is any network interruption or data issues, the app works smoothly and effectively. Thus, helping the user in safe driving through hotspot regions.

7.2 Efficiency of the App For the time being to test the efficiency of the app, a few test drives were conducted among a few young drivers. It worked satisfactorily. The mobile phone is an essential device nowadays and the beep sound from the device alerts the driver in the most efficient way. Usually, the reaction time of the driver is 2 s, the warning in advance of 200 m helps the diver to be more careful while approaching the hotspot region.

8 Results and Discussion A detailed analysis of the characteristics of the accidents that occurred based on the reported accidents and a detailed investigation of the hotspots was done on the study stretch and the following conclusions were made. • Number of accident cases increases as the number of registered vehicles increases day by day. Road safety measures implemented so far are not sufficient to prevent fatalities or loss of life due to road accidents • 65% of the accidents were reported during the daytime as the number of vehicles was more in the daytime, 15% were after 9 pm even with less traffic volume. • 55% of vehicles involved in accidents were two-wheelers and young male drivers with age group 18–29 were involved in more accidents. • 74% of the accidents were severe with grievous injury or loss of life and more cases happened in sunny clear climates. • 27 hotspots were identified in the stretch. From the detailed investigation of geometrical and traffic features most of the accidents occurred in straight stretches, and curves are due to overspeed, carelessness of the young drivers, absence of warning signs insufficient carriageway and shoulder width or no shoulders. • To reduce the number of crashes due to overspeeding of young drivers, absence of warning signs an Android mobile application ‘SAAVADHAN’ was developed which alerts the driver 200 m in advance of the hotspot region by a beep sound. The working of the application was checked and found satisfactory. • From local interactions and also from interaction with police officials, a number of minor accidents were unreported and the accident data from the police records alone are not sufficient to do a comprehensive analysis.

Comprehensive Analysis of Road Accidents and Surrogate Measures …

581

• The accidents due to driver negligence can be reduced by giving road safety education and through enforcement measures. Accidents due to overspeed can be reduced by providing optical treatments which can be applied to the road surface as markings, which creates a feeling or even illusion to the driver that he is going too fast.

References 1. Bener A, Yildirim E, eOzkan T, Lajunen T (2017) Driver sleepiness, fatigue, careless behavior and risk of motor vehicle crash and injury: Population based case and control study. J Traffic Transp Eng 4:496–502 2. Bian H, Wang L, Zhao W, Liu H (2013) Study on driver stress response ability on rural roads in China. In: Proceedings of 2nd International conference on transporation information and safety, pp 1287–1297 3. Baktha K (2017) Mobile application development: All the steps and guidelines for successful creation of mobile app: Case study. Int J Comput Sci Mob Comput 6(9):15–20 4. Castelloa Dl, Findleyb DJ, Torregrosaa FJC, Garciaa A (2019) Calibration of inertial consistency models on North Carolina two-lane rural roads. Accid Anal Prevention 27:236–245 5. Chakrabartya N, Guptab K (2013) Analysis of driver behaviour and crash characteristics during adverse weather conditions. In: 2nd Conference of transportation research group of india. Procedia–social and behavioral sciences, vol 104. pp 1048–1057 6. Halla T, Tarkob AP (2019) Adequacy of negative binomial models for managing safety on rural local roads. Accid Anal Prevention 128:148–158 7. Hu S, Xiang O (2012) Characteristic analysis of traffic accidents on rural roads. In: Proceedings of the 12th international conference of transportation professionals, pp 2506–2513 8. Rolison J, Regev S, Moutari S, Feeney A (2018) What are the factors that contribute to road accidents? An assessment of law enforcement views, ordinary drivers’ opinions, and road accident records. Accid Anal Prev 115:11–24 9. Vayalamkuzhi P, Amirthalingam V (2014) Influence of geometric design characteristics on under heterogeneous traffic flow. J Traffic Transp Eng 3:559–570 10. Vayalamkuzhi P, Amirthalingam V (2016) Development of comprehensive crash models for four-lane divided highways in heterogeneous traffic condition. In: 11th Transportation planning and implementation methodologies for developing countries. transportation research procedia, vol 17, pp 626–635 11. Wang Y, Yao Y (2015) Research on crash prediction and safety performance of two-way two-lane rural roads in Gelderland. In: Proceedings of the 152th international conference of transportation professionals, pp 3071–3080 12. Wu L, Chu J, Ci Y, Feng S, Liu X (2015) Engineering solutions to enhance traffic safety performance on two-lane highways, vol 2015. Hindawi Publishing Corporation, p 7

Visualising Blackspot Improvement at Nagpur Raghav Chawla, S. Velmurugan, Mukti Advani, and K. Ravinder

Abstract Nagpur, the “Orange City” of the country is situated in the state of Maharashtra. Nagpur is one of the fastest economically growing cities and part of the “Smart City Mission”. In this paper, the city of Nagpur has been identified as the test bed to understand and demonstrate the before and after scenarios of black spot improvement through 3-D modelling. In this regard, road crash data obtained from Nagpur Police in the form of First Information Reports (FIRs) was analysed to identify the black spots conforming to the protocol of the Ministry of Road Transport and Highways (MoRT&H). The geometric countermeasures deduced for two out of the 37 locations are illustrated by presenting ‘before’ and ‘after’ scenarios through 3-D Modelling with the primary objective of pictorially depicting the efficacy of the proposed intervention for two out of the 37 locations in Nagpur. The above form of pictorial illustration of the detailed Geometric Design Plan (GDP) for the above black spots will serve as an eye-opener for the relevant stakeholders in terms of undertaking steps towards the implementation of the suggested black spot remedial measures. It is clarified that the overall project termed Intelligent Solutions for Road Safety through Technology and Engineering (iRASTE) includes redesigning the road geometry as well as applying Artificial Intelligence (AI) technology to achieve its aim of reducing road fatalities by at least 50%, safety of pedestrians, and ease in traffic flow. However, this paper aims to highlight the importance of 3-D visualization for the representation of design interventions through which the policymakers, stakeholders and the general public alike can easily understand. Keywords Transportation design · Road safety · Crash prevention · 3-D Visualization

R. Chawla Manipal School of Architecture and Planning, Manipal Academy of Higher Education, Udupi, ManipalKarnataka 576104, India e-mail: [email protected] S. Velmurugan (B) · M. Advani · K. Ravinder CSIR-Central Road Research Institute, 110025, Mathura Road, New Delhi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_46

583

584

R. Chawla et al.

1 Preamble In any city, all roads are safe until they are not, hence proper road designs are required. Road geometry plays an important role in ensuring the safety of all road users on the road network. To address the above, the city of Nagpur has been identified as the test bed to understand the demonstrate the ‘before’ and ‘after’ scenarios of a detailed geometric design plan (GDP) through 3-dimensional (3-D) Modelling in this paper. In this regard, road crash data obtained from Nagpur Police in the form of First Information Reports (FIRs) was analysed to identify the black spots conforming to the protocol of the Ministry of Road Transport and Highways (MoRT&H). While designing the above improvement plans at the identified black spots, the modal share was taken on board along with the functional classification of the road links in the form of arterial/sub-arterial/collector which carry huge volumes of traffic [1]. Further, the lanes leading to the markets are very narrow and also have unregulated parking which causes traffic congestion majorly during peak hours. Lack of proper road and street designs causes traffic-related issues which not only include congestion problems but also can lead to road crashes. Considering the above, it is envisaged to develop remedial measures, i.e. countermeasures for the identified black spots. In this context, 37 Black spots have been identified based on the analysis of the First Information Report (FIR) obtained from the Nagpur Police. After this, the geometric countermeasures are deduced for the above in Nagpur and two out of the 37 locations are illustrated by presenting ‘before’ and ‘after’ scenarios through 3-D Modelling with the primary objective of pictorially depicting the efficacy of the proposed intervention. The above form of pictorial illustration of the detailed Geometric Design Plan (GDP) for the above black spots will serve as an eye opener for the relevant stakeholders in terms of undertaking steps towards the implementation of the suggested black spot remedial measures. It shall be noted that under the larger umbrella of the study titled, Project: iRASTE, i.e. Intelligent Solution for Road Safety through Technology and Engineering the above-identified blackspots i.e. crash-prone areas with a history of fatalities and several crashes as well as the identification of the grey spots i.e. locations vulnerable to becoming black spots have been identified and presented in Fig. 1. However, in this paper, the engineering measures conceived for a couple of black spots have been illustrated through 3-D Modelling, and hence any discussion on the grey spots is beyond the purview of this paper.

2 An Overview of 3-D Modelling 3-D visualization is a decade-old technology/tool. However, this technology kept evolving to become hyper-realistic in representing anything in a three-dimensional form on a screen or paper and even in physical forms. The need for the most advanced 3-D visualization came from one simple concept–bridging communication gaps between clients and artists. One of the primordial purposes of visualization

Visualising Blackspot Improvement at Nagpur

585

Fig. 1 Black spot locations (Source iRASTE Dashboard)

was communication. Because of that, visualization has continuously been evolving”. [2]The very first 3-D modeling software was developed by Iwan Edward Sutherland at MIT in 1958 which revolutionized the human–computer interaction. Since then, a lot of 3-D modelling software has been developed. This method of visualization soon found its way into the architecture industry and then later into the film industry. The use of 3-D Modelling is also evident in developing a greater understanding of complicated designs in a much nontechnical manner for a layman to understand. In India mostly, 3-D modelling and rendering of walkthroughs are limited to architectural firms and some government departments, which require 3-D visualizations. Many large-scale projects such as highway design or a redesign of a long stretch of the street including intersections are mostly done in the 2-D format in Computer Aided Design (CAD) software, they hardly find their way to 3-D modelling and rendering. One of the most popular projects which used 3-D rendered walkthroughs is the Coastal Road Project in Mumbai, the video of the conceptual design circulated on various social media platforms made it very easy for even the general public to understand the proposal. In this regard, the study report of the Federal Highway Administration (FHA), Washington DC [3] lists the benefits that have been derived out of the 3-D visualization accomplished through the 3-D illustration of the benefits that accrued through the depiction of Road Safety Audit (RSA) projects: • It enabled members of the RSA team to read the design more efficiently, especially those who are not well-versed in reading complex designs. • The model helped in depicting all the details of the design both the horizontal features as well as the vertical ones by giving the idea of the surroundings, abutting land uses, heights of the buildings, etc. • 3-D model provided a platform in the form of vantage points from which the proposal can be viewed as compared to 2-D.

586

R. Chawla et al.

Fig. 2 Study methodology

• It helped in easily conveying design to all the stakeholders which include nontechnical persons. • In Virginia, the 3-D model helped visualize the impacts of proposed designs in sensitive areas.” • In Rhode Island, the 3-D model allowed for the review of locations that were not readily accessible to the RSA team [3]. Considering the above inherent benefits of the pictorial illustration of the 3-D Modelling, an effort has been made in this paper to illustrate the Detailed Geometric Design Plan (GDP) conceived for two out of the 37 black spots in the city of Nagpur through the above tool.

3 Methodology The study Methodology devised to demonstrate the conceived GDP is presented in Fig. 2. The 3-D models are developed for the existing and the proposed design by making a walk-through video by placing them side by side. These depict the interventions, some of which are also captured in still renders as shown in the succeeding sections.

4 Visualization of GDP Conceived for Black Spots As mentioned earlier, the detailed geometric plan (DGP) deduced for two out of the 37 locations is illustrated by presenting ‘before’ and ‘after’ scenarios through 3-D Modelling is illustrated in this paper. This focuses on one Intersection (namely Jhansi Rani Square having five arms) and one midblock (namely, the road stretch from Telephone Exchange to CA road) in Nagpur city., Nagpur. The data deployed

Visualising Blackspot Improvement at Nagpur

587

for 3-D modelling and walkthrough rendering for the two black spots encompassed 2-D CAD drawing depicting the existing physical survey plan and the proposed detailed GDP, Classified Volume Counts, Pedestrian Volume Counts, Stationary and mobile Videos, Google Earth, various [4–6] and Detailed Project Report (DPR) of Nagpur Metro.

4.1 Jhansi Rani Square 1 This intersection is located near the Sitabuldi area in Nagpur. It has five arms consisting of -AH-46, SH-255, and SH266, a road leading to Nagpur Railway Station, and one to Sitabuldi Metro station. The traffic count observed for a period of 12 h in both directions is 55,774 PCUs (74,871 vehicles) with an average speed of 31.9 kmph. The most preferred mode of transport in this area is two-wheelers (62%), and the overall modal split at this intersection is shown in Fig. 3. The average crashes at this location are around 32 before countermeasures and the number is estimated to decrease to 11 after the design intervention (which includes a host of countermeasures, traffic calming measures, etc.) which is a significant 66% decrease. The existing road geometrics is depicted in Figs. 4 and 5 whereas Fig. 6 presents the existing road geometry in the form of 2-D CAD drawings illustrating the existing situation. The proposed intervention is depicted in Fig. 7 which broadly encompasses the following: • Channelizing islands and depressed medians provision with cement concrete bollards to provide safe pedestrian crossing.

Fig. 3 Modal split at Jhansi rani square 1

588 Fig. 4 Absence of zebra crossing

Fig. 5 Absence of box marking at the crash prone location

Fig. 6 Existing geometric profile of the candidate intersection

R. Chawla et al.

Visualising Blackspot Improvement at Nagpur

589

Fig. 7 Proposed geometric design plan of the candidate intersection in 2-Dform

• Pedestrian crossing signals are to be installed at a distance of 45 m from the zebra crossing; the signal will be open for pedestrians for 20 s during the 60-s signal time. • Transverse Bar Makings (TBMs) and rumble strips to be provided in both directions of travel. Provision of missing informatory road signs, retro-reflective pavement marking, edge delineators, and road studs. • Provision of a speed table equipped with Blinking Solar Signals at free left turns on every approach. Figure 8 presents the typical snapshot illustrating the existing conditions in the case of the Do Nothing, i.e. Business As Usual (BAU) scenario taken from the 3-D rendering video. On the other hand, Fig. 9 presents a bird eye view with interventions and similarly, Fig. 10 presents the typical view from the driver’s seat depicting the proposed interventions whereas Fig. 11 presents the typical view of safety measures for one of the intersection approaches after interventions. Basically, the 3-D rendering video has been created by covering all 5 arms of the candidate intersection spanning a length of 250 m from the center of the intersection. As mentioned earlier, this kind of rendering thus helps in showcasing the ‘before’ and ‘after’ scenarios.

590

R. Chawla et al.

Fig. 8 Typical illustration of 3-D rendering of the ground conditions without intervention

Fig. 9 Typical illustration of 3-D rendering of the ground conditions with intervention Fig. 10 3-D view from the driver’s seat with intervention

Visualising Blackspot Improvement at Nagpur

591

Fig. 11 Typical 3-D view of safety measures on one of the intersection approaches after intervention

4.2 Telephone Exchange to CA Road This is a mid-block location (on AH-46) which is located between two intersectionsnamely, Telephone Exchange Square and Chappru Nagar Square. The proposal spans a distance of about 1 km. This black spot intervention comprised the redesign of the entire stretch of the road including the above intersections. The traffic count observed at the midblock section during the 12-h covering both directions of travel is 40,835 PCUs (50,557 vehicles) with an average speed of 39.5 km/h. The most preferred mode of transport in this section is again two-wheelers (69%), and the overall modal split at this intersection is shown in Fig. 12. The average crashes at this location are around 37 before countermeasures and the number is estimated to decrease to 15 in the vent of the implementation of the above design intervention (like DGP, traffic calming measures, etc.) which is a significant 59% decrease. The existing traffic scenario and road geometrics at this midblock are presented in Figs. 13 and 14 respectively whereas Fig. 15 presents the existing road geometry in the form of 2-D CAD drawings illustrating the existing situation. Figures 16 and 17 present the typical snapshot illustrating the existing conditions in the case of BAU scenario taken from the 3-D rendering video. The proposed intervention is depicted in Fig. 18 which broadly encompasses the following: • Bus bay provision on both sides of the highway after the intersection approach in Chappru Chowk. • Channelizing the island with depressed medians to allow safe pedestrian crossing at both Chappru Chowk and Dr. Rajendra Prasad Chowk. • Incorporation of traffic calming measures before reaching the Telephone Exchange metro station to reduce the travelling speeds where pedestrian volume is on the higher side.

592

R. Chawla et al.

Fig. 12 Modal split on telephone exchange to CA Road

(a)

(b)

Fig. 13 a Absence of zebra crossing at the Rajendra Chowk b On street parking on the corridor: Major safety issue

Fig. 14 Existing geometric profile of the candidate Midblock location

Visualising Blackspot Improvement at Nagpur

593

Fig. 15 Proposed geometric design plan of the candidate Midblock location spanning for 1 km in 2-D Form

(a)

(b)

Fig. 16 a & b Typical illustration of 3-D rendering of the ground conditions without intervention on the candidate Midblock

Fig. 17 Typical illustration of 3-D rendering of the ground conditions without intervention on the candidate midblock at Rajendra Chowk

The images above depict the change in the level of understanding and interpretation of the proposal. The 2-D drawings consist of the technical data from the survey by using the proposed GDP drawings whereas the 3-D renders translate them into a realistic form that makes them understandable for the stakeholders and even the common public.

594

R. Chawla et al.

Fig. 18 Typical illustration of 3-D rendering of the ground conditions with intervention on the candidate intersection at Rajendra Chowk

5 Concluding Remarks and Future Scope This study demonstrated the effectiveness of the 3-D model and its rendering helped in understanding its efficacy. Thus it can be inferred that a 3-D walkthrough can let the observer move around the road or the entire stretch and get a realistic idea of the proposal and thus provide angles that cannot not be accessed on foot due to running traffic and other safety concerns. Some inherent benefits derived through the study are listed: • A 3-D walkthrough or even a still render gives the z-axis -the missing vertical which adds the height factor to the drawing and hence makes it easier to understand in terms of scale, details of the location and impact. • This type of visualization can open the door to public participation as well, since this depicts the design in a realistic format, it becomes easy for the policymakers and stakeholders to better understand the proposal and thus give feedback or suggestions. • Further advances in this field can also lead to the fusion of AR (Augmented Reality) and AI (Artificial Intelligence) into 3-D to predict and display future conflict points by using the CVC data and future projections. In summary, this can be reckoned as an efficient tool for road development projects just like in other industries to showcase the design in a hyper-realistic format.

References 1. CRRI (2022) iRaste Interim report 2. How 3D visualization developed and got where it is today. https://www.easyrender.com/a/how3d-visualization-developed-and-got-where-it-is-today (Accessed 23 Aug 2023)

Visualising Blackspot Improvement at Nagpur

595

3. Nabors D, Soika J (2013) Road safety audit case studies: Using three-dimensional design visualization in the road safety audit process 4. IRC:67 (2022) Code of practice for road signs 5. IRC:79 (2019) Recommended practice for road delineators 6. IRC:35 (2015) Code of practice for road markings

Proactive Safety Assessment at Unsignalized T-Intersection Using Surrogate Safety Measures: A Case Study of Bhopal City C. Noor Mohammed Parvez , Pritikana Das, and Dungar Singh

Abstract Every year, road accidents cost billions of dollars and damage millions of people around the world. Among all critical points, intersections pose special safety concerns because of the high probability of critical conflicts resulting from unsafe driver actions and maneuvers. As a result, assessing road safety, particularly at uncontrolled T-intersections in mixed traffic, becomes critical. Over the last few decades, Surrogate Safety Measures (SSM) have attracted a lot of study attention for analyzing road safety issues. These approaches are designed to be proactive, that do not rely on collisions, and require shorter observation time periods to create acceptable safety assessments. In this study, two uncontrolled intersections in Bhopal City were chosen as research areas for safety evaluation by adopting the Surrogate Safety Assessment approach. The study focused on crossing conflicts among right-turning vehicles and through traffic as they are considered severe among other conflicts. Post-Encroachment Time (PET) and conflicting vehicle speed of through traffic are used to determine critical conflicts. Further, Encroachment Time (ET) is taken as surrogate indicators to identify severity levels of right turning movements using the clustering technique. Keywords Surrogate safety measures · PET · Critical conflict · Severity level

1 Introduction Road safety has always been measured using accident data, which is essentially a reactive technique, despite the fact that this method has time and efficiency limitations. According to past literatures on traffic conflict techniques, employing surrogate safety data provides for a faster evolution of safety than using long-term accident data. At any given uncontrolled intersection, a large number of vehicle-to-vehicle and vehicle-to-pedestrian interactions could result n a crash. Road intersections are C. N. M. Parvez (B) · P. Das · D. Singh Maulana Azad National Institute of Technology Bhopal, Bhopal 462003, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_47

597

598

C. N. M. Parvez et al.

traffic merging points and hence are prone to accidents [7]. Within the intersection category, T junction accounts for the largest share of accidents, persons killed and injured [7]. As a result, assessing current road safety indications, particularly at unsignalized T-intersections in mixed traffic, becomes critical. The majority of traffic safety assessments are based on the analysis of historical accident data, which is reactive in character; it’s as if they’re waiting for a road accident to happen before implementing their remedies. Furthermore, researchers suggested a proactive approach based on surrogate safety measures (SSMs) as a new method for assessing collisions at signalized and non-signalized junctions. The key benefit of this method is that it can help forecast the frequency of an upcoming road crash due to poor road geometry caused by the aforementioned variables, therefore serving as a more efficient and reliable proximate measure of traffic safety. This study focuses on proactive safety evaluation of T-intersections in Bhopal city. Suitable SSM parameters have been identified from the literature to assess the safety at uncontrolled Intersections for right-turning conflicts. Surrogate safety indicators namely Post-Encroachment Time (PET), Conflicting Speed, and Encroachment Time (ET) are considered in this study. Percentage of critical conflicts observed based on PET and conflicting speed. Severity levels of safety measures have been defined using the clustering technique. The study outcomes will be useful for field engineers, planners and decision-makers to understand the present scenario and to provide safety appropriate safety measures. Definitions and Terminology 1. Post-Encroachment Time (PET): Time-lapse between the end of encroachment of the turning vehicle and the time that the through vehicle actually arrives at the potential point of collision. The value is recorded in seconds [1]. 2. Conflicting Speed: Speed of the through vehicle involved in the conflict event with a right-turning vehicle. 3. Encroachment Time (ET): Time duration during which the turning vehicle infringes upon the right-of-way of the through vehicle. This value is expressed second Allen [1].

2 Literature Review In a country with a large population, such as India, traffic safety is still a reactive strategy. However, such an analysis is typically performed as an “after-thought” rather than proactively. Safety evaluation has historically been based on policereported crash data in order to decrease crashes. The analysis of traffic crash data can be useful to understand the general pattern of crash occurrence and to identify the primary contributory variables that can be useful in implementing necessary countermeasures. However, in addition to the other limitations associated with the traditional technique, the non-availability of accident data and inaccurate information about the crash pattern and location are all too typical in developing countries. To

Proactive Safety Assessment at Unsignalized T-Intersection Using …

599

overcome this drawback, some studies have suggested the use of short-term, indirect traffic safety measures that are ‘proactive’ in character and can be utilized instead of historical collision data for a more reliable and faster safety assessment [1]. As a result, a surrogate approach known as “conflict analysis” was created to overcome the paucity of previous accident and exposure data. In a majority of research, the PET threshold was used, either at random or by taking into account the perception-reaction time associated with Stopping Sight Distance (SSD). Babu and Vedagiri [12] in their study recorded PET values and the speeds of their related conflicting through traffic to observe conflicts. Taking mixed traffic into account, Critical conflicts were discovered by Paul and Ghosh [8], Babu and Vedagiri [13] utilized critical speed, which was established using two surrogate safety indicators, speed of conflicting vehicles and PET. Pawar et al. [4] used PET as a surrogate safety measure to investigate the impact of intersection control measures (SSM). By collecting data from ten crossings, Paul and Ghosh [9] estimated a suitable PET threshold for classifying critical conflicts in a highly heterogeneous traffic environment. And correlated PET values with Crash Data for considering different classes of vehicles and observed PET threshold is 1 s. The intersections were ranked based on the cumulative number of PETs and accompanying crash data. The critical speed parameter, defined by Paul and Ghosh [8], is derived using the idea of braking distance and is used to identify important conflicts. It was discovered that average PET values at various locations range between 2.13 and 3.16 s. Larger proportions of conflicts that are critical were observed when with right-turning HVs. The Required Deceleration Rate for every conflict is determined as v/2PET from braking distance. Based on these critical conflicts, PET and RDR are determined and ranked 3 T-Intersection. Babu and Vedagiri [13] proposed critical speed to identify critical conflicts which are calculated based on the braking distance concept for the particular critical value. They found that right-turning light motorized vehicles (LMVs) such as auto-rickshaws, cars, and minibuses are more vulnerable than large vehicles (buses and trucks) and 2W. Babu and Vedagiri [12] identify important conflicts, with the term critical speed has been proposed. The braking distance concept is used to establish the critical speed for a certain PET value. There is a substantial percentage of observed confrontations that are critical at the crossing, according to their findings. This shows that right-turning vehicle drivers are willing to take chances and accept smaller gaps in through traffic.

3 Study Methodology The methodology of a research project outlines the complete work process and plan for achieving the research project’s goal. It is a methodical process to achieve the goal of this study. The adopted methodology for the study is shown in Fig. 1. Various surrogate safety measures have been identified from past studies, utilized and tested. For this study, Post-encroachment Time, Difference in conflicting velocities, and Encroachment time were considered for safety assessment. PET is a quantitative

600

C. N. M. Parvez et al.

Fig. 1 Methodology flowchart

method for determining the state of a conflict. Because traffic indicators such as stop and yield signs are frequently absent at uncontrolled intersections, drivers have little control over their approaching speed. At intersections, the higher approaching speed of vehicles contributes to the severity of conflict that results in a collision. Because PET alone cannot judge, severity of a conflict, the approaching speed of conflicting through traffic, is considered in determining the severity and frequency of a conflict and evaluating the intersection’s safety. The extracted data is used for analysis from which critical conflicts and clustering were obtained.

4 Data Collection and Extraction The traffic data was obtained using a videography technique on a working day in October 2021 under fair weather conditions at the subject’s un-signalized Tintersections. Road inventory and traffic volume details of selected intersections are also collected and shown in Table.1. Two uncontrolled intersections were selected based on having different geometric and traffic characteristics, including the presence of high-rise buildings near the location to capture data effectively, variable traffic demand at a different site to get more variations in safety indicators, vehicles travel at desired speed with less obstruction to flow, having both commercial and residential land use. Further, to track the movement of turning vehicles, the conflict area is divided into grids of 3.5 m × 3.5 m squares, with a lane width of 3.5 m, then overlaid with Kinovea software. Data extraction of 1 h (10 am−11 am) has been performed at selected study locations to evaluate safety in this study. The time delay between the offending vehicle (turning vehicle) leaving the conflict grid and the

Proactive Safety Assessment at Unsignalized T-Intersection Using …

601

Table 1 Road inventory and volume details at two intersections Study location

No. of lanes in approaches and departures

Width of approaches/ departures (m)

Traffic volume (veh/hr)

Major

Minor

Major

Minor

Through

Right turn

Left turn

Ratnagiri Tiraha

6

6

3.5

3.5

2311

1126

512

Neelbad Tiraha

6

2

3

3.5

1609

477

493

conflicting vehicle (opposite through vehicle) entering the respective conflict grid is used to calculate PET values. The speeds of conflicting vehicles are also calculated by recording the time it takes to travel the distance between grids (three to four grids). PET values and the speeds of related conflicting vehicles along with the type of right turning and conflicting vehicles are noted. Similarly, ET is also extracted from video using Kinovea.

5 Data Analysis 5.1 General PET and conflicting speeds of through traffic are used to determine critical conflicts for right turning and through traffic conflicts. Further, Encroachment Time (ET) are taken as surrogate indicators to identify the severity level of through and right turning movements respectively using the clustering technique.

5.2 Descriptive Statistics The following Table 2 shows the descriptive statistics of the extracted data at three study locations namely Ratnagiri Tiraha, and Neelbad Tiraha. The aggregate mean, standard deviation, variance of the SSM indicators evaluated to assess the safety of 3-legged uncontrolled crossings in Bhopal City are shown in the descriptive statistics.

5.3 Distributions of Conflicts for All Right-Turning Vehicles with Through Traffic For a conflict with such PET value and speed of the conflicting vehicle, when rightturning vehicle just left the conflict area, the conflicting through vehicle is at a distance

602

C. N. M. Parvez et al.

Table 2 Descriptive Parameters of SSM indicators at study locations Parameter

N

Range

Mean

Standard deviation

Variance

Ratnagiri Tiraha PET

896

10.40

1.432

1.217

1.481

Conflicting speed

896

42.038

18.433

6.911

47.758

ET

1062

33.050

6.153

3.482

12.123

Neelbad Tiraha PET

404

18.00

4.07

3.47

15.546

Conflicting speed

404

39.586

20.329

6.571

43.178

ET

396

29.400

4.283

3.521

12.396

Note N indicates Number of observations

equal to PET times the conflicting vehicle’s speed (PET × conflicting vehicle’s speed). The conflict is not critical if this distance exceeds the stopping distance required for the conflicting vehicle’s speed. Conflict is critical if the distance is shorter than the stopping distance required. To distinguish between critical and non-critical conflicts, the distance available is equated to braking distance [12]. Because opposing vehicle drivers have already reacted to a crossing manoeuvre, the perception of distance was neglected. The formula v2/2gf is used to compute the braking distance ‘d,’ where v is the opposing vehicle’s speed in m/s, g is gravity acceleration in m/s2 , and f is the coefficient of friction between the road surface and tyre. The critical speed for that PET value is computed using PET* 2 gf, which is calculated by multiplying the available distance by the braking distance. Using this method, critical speeds for specific PET levels are computed using g = 9.81 m/s2 , and coefficient of friction = 0.35. Available distance = Braking distance V x PET = V2 /2gf Critical speed, V = 2gf × PET In the current study, this concept was applied to assess the safety of an unsignalized intersection. To detect traffic conflicts, PET levels and speeds of linked conflicting through vehicles are utilized. Critical conflicts are calculated using the PET value, associated critical speed, and speeds of conflicting vehicles. If the conflicting speed is greater than the matching critical speed for a conflict with a defined PET value, it is a critical conflict. The PET value of each conflict, as well as the speed of the opposing car, influence the outcome. This study considers PET values higher than 0 s and less than 6 s since many recorded conflicts with negative PET readings have slower speeds. At two crossings with PET ranges ranging from 0 to 6 s, a total of 896, and 404 conflicts at

Proactive Safety Assessment at Unsignalized T-Intersection Using …

603

Ratnagiri Tiraha, and Neelbad Tiraha were discovered. As indicated in Tables 3 and 4, the observed conflicts are segregated based on PET values that are categorized into 12 categories with a 0.5 s increment. The Lower Limit (LL) of PET for each of these categories is used to identify critical speeds. The distribution of these conflicts is also given in Tables 3 and 4, which is separated by the type of right-turning vehicle. A brief summary of critical conflicts of critical conflicts at study sites is shown in Table 5. A total of 47.3%, and 19.06% of conflicts were found critical at Ratnagiri Tiraha, and Neelbad Tiraha, respectively. At Ratnagiri, Tiraha conflicts involving cars are found to be at a higher risk with 51.2% conflicts involving cars being critical. Whereas, at Neelbad Tiraha Two-wheelers are at higher risk with 24.66% conflicts involving Two-wheelers being critical.

5.4 Determination of Severity Levels Using Clustering Technique Encroachment Time (ET) are chosen as safety indicators to define severity levels of right turning two intersections. Encroachment time was extracted from video-graphic data using Kinovea software as the time difference between entering and leaving time stamps in through traffic path by right-turning vehicles (as given in Eq. 1). Which indicates time spent by right-turning traffic in through traffic path [1]. ET(Rt) = texit (Rt) − tentry (Rt)

(1)

where ET(Rt) is Encroachment of Right turning (Rt) vehicle. The severity of probable road crashes will increase with an increase in the value of ET for right turning. In order to classify the severity of probable road collisions based on the ET, all the values ET are grouped using the clustering technique Cluster analysis is the process of categorizing items based on data in the data set that describes their relationships. In this study, the K-means clustering is used as a clustering technique. K-means clustering is a well-known hard partitioning approach that is particularly useful for forming small clusters from large datasets. The k-means function divides the observed data into k mutually exclusive clusters and provides a vector of indices that indicate which of the k clusters each observation belongs to. After classifying the data, the silhouette index was used to validate the results of each clustering technique. The cluster analysis and validation were carried out using MATLAB software. Finally, the range and threshold values of 3 clusters using the k-means clustering technique for ET are presented in Table 6. The last column of the table shows average silhouette index for all clusters. Further, three severity levels are developed based on obtained threshold values from K-mean clustering as Severity Level (SL) A, B, and C which denote Safe, Moderately Safe, and Unsafe conditions of road users, respectively, as shown in Table 7. The frequency of probable road crashes for unsafe

111.2

123.6

135.9

5

5.5

6

4.5

5

5.5

86.5

98.9

4

4.5

3.5

74.2

61.8

4

3.5

3

49.4

2.5

3

2

2.5

37.1

2

1.5

12.4

24.7

1

1.5

0.5

0

0.5

0

UL

LL

1

Critical speed kmph

PET (sec)

0.45

0.67

0.89

2.91

2.80

4.59

9.96

10.51

20.25

28.30

17.00

0.00 26.06

0.45 85.91

47.32

0.45

0.67

0.78

2.68

2.46

3.69

7.83

9.28

16.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.22

3.58

24.61

18.90

0.00

15.55

Two-Wheeler Conflicts (%)

Critical conflicts (%)

Ratnagiri Tiraha

Conflicts (%)

41.28

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.11

3.47

22.15

15.55

Critical conflicts (%)

12.53

0.00

0.00

0.11

0.00

0.45

0.45

0.67

0.89

0.78

3.47

3.80

1.90

Conflicts (%)

5.48

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.22

3.36

1.90

Critical conflicts (%)

Three-Wheeler

Table 3 Distributions of conflicts for right-turning with through traffic at Ratnagiri Tiraha LCV

13.87

0.34

0.11

0.00

0.00

0.00

0.34

0.45

1.34

1.45

4.36

2.35

3.13

Conflicts (%)

5.37

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

2.24

3.13

Critical conflicts (%)

Car

43.62

0.11

0.34

0.34

0.22

0.89

0.78

1.68

4.14

4.25

13.09

10.51

7.27

Conflicts (%)

22.26

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

1.12

5.37

8.50

7.27

Critical conflicts (%)

604 C. N. M. Parvez et al.

86.5

98.9

111.2

123.6

135.9

1.5

2

2.5

3

3.5

4

4.5

5

5.5

6

1

1.5

2

2.5

3

3.5

4

4.5

5

5.5

0.00

1

4.21

4.21

3.96

7.43

7.18

6.19

7.43

7.67

8.17

8.17

9.90

0.00

2.97 72.28

19.06

3.47

4.46

2.97

7.67

6.68

6.19

5.69

7.67

7.67

8.17

8.66

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.25

0.99

7.92

9.90

0.00

17.83

0.00

0.01

0.00

0.00

0.00

0.00

0.00

0.00

0.25

0.99

7.92

8.66

Critical conflicts (%)

Conflicts (%)

Conflicts (%)

Critical conflicts (%)

Two-Wheeler

Neelbad Tiraha

16.34

0.74

0.25

0.50

0.25

0.50

0.50

0.74

1.24

0.74

8.17

0.74

1.98

Conflicts (%)

3.223

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.50

0.74

1.98

Critical conflicts (%)

Three-wheeler

Where, UL = Upper Limit of PET, LL = Lower Limit of PET, and LCV = Light Commercial Vehicle

74.2

61.8

49.4

37.1

24.7

12.4

0.5

0

UL

Critical speed kmph

0.5

LL

PET (sec)

Table 4 Distributions of conflicts for right-turning with through traffic at Neelbad Tiraha

15.8

0.00

0.50

0.25

0.50

0.74

0.50

0.25

0.99

1.49

8.42

0.50

1.73

Conflicts (%)

LCV

2.23

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.50

1.73

Critical conflicts (%)

32.92

0.99

0.99

0.74

0.25

1.24

1.73

2.23

1.49

1.24

16.83

2.72

2.48

Conflicts (%)

Car

4.21

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.00

0.25

1.49

2.48

Critical conflicts (%)

Proactive Safety Assessment at Unsignalized T-Intersection Using … 605

606

C. N. M. Parvez et al.

Table 5 Brief summary of critical conflicts Study location Ratnagiri Tiraha

Neelbad Tiraha

Conflicts

Total

Two wheelers

Three wheelers

LCV

Car

896

770

112

124

391

Critical conflicts

424

370

49

48

200

Percent of critical conflicts in total conflicts

47.3

48.05

43.75

38.7

51.2

Conflicts

404

292

66

64

133

Critical conflicts

77

72

13

9

17

Percent of critical conflicts in total conflicts

19.06

24.66

19.70

14.1

12.8

cases based on their ET for selected intersections using K-means is presented in Table 8. The results from Table 8 show that the percentage of unsafe right-turn vehicles is higher at Ratnagiri Tiraha with 5.65% whereas at Neelbad Tiraha is 1.52%. So, the probability of crashes involving right-turning vehicles may be seen more at the former than at the later. Table 6 Results of K-means clustering along with Silhouette Index Study location

Parameter

Range

Thresholds

Min

Max

Ratnagiri Tiraha

ET

0.80

33.85

6.525, 12.5

Neelbad Tiraha

ET

0.48

29.88

5.68, 16.14

Table 7 Classification of encroachment time (in sec) for severity of right turning traffic Study location

Threshold Values for ET using K-means clustering

Severity level

Safety of right-turning movement

Ratnagiri Tiraha

< 6.525

A

Safe

Neelbad Tiraha

6.525–12.5

B

Moderately safe

>12.5

C

Unsafe

< 5.68

A

Safe

5.68–16.14

B

Moderately safe

>16.14

C

Unsafe

Table 8 Percentage of vehicles fall under the unsafe case, i.e. when SL = C

Parameter

Ratnagiri Tiraha

Neelbad Tiraha

ET

5.65

1.52

Proactive Safety Assessment at Unsignalized T-Intersection Using …

607

6 Conclusion This provided a study that developed a methodology for evaluating a surrogate safety indicator, PET, and conflicting speed to measure traffic safety at three-arm uncontrolled intersections. Further, severity levels to assess the safety right turning movement using ET. PET threshold values are used to determine critical conflicts. However, at intersections with mixed traffic and varying speeds, relying solely on PET to assess safety is insufficient. As a result, conflicts are observed in the current study employing two surrogate indicators PET, and related conflicting vehicle’s speed. To determine critical conflicts at intersections, the critical speed is recommended. The braking distance concept is used to find out the critical speed for a certain PET value. There is a substantial percentage of observed conflicts at the intersection are critical, according to the findings. This demonstrates that right-turning vehicle drivers are willing to take chances and accept short gaps in through traffic paths, which is unsafe. At Ratnagiri Tiraha conflicts involving cars are found to be at a higher risk with 51.2% of conflicts involving cars being critical. Whereas, at Neelbad Tiraha Two-wheelers are at higher risk with 24.7% of conflicts involving Two-wheelers being critical. This could be owing to their high proportion of volumes and high speed. Further, namely Encroachment Time (ET) is used to define severity levels of right turning vehicles. Three unsafe severity levels were developed namely less unsafe, moderately unsafe, and highly unsafe using the K-means clustering technique. Among two intersections Ratnagiri Tiraha found it unsafe for right-turning traffic with 5.65% of vehicles falling under severity Level C, and at Neelbad Tiraha it is observed as 1.52% of vehicles falling under severity level C. The proposed approach provides a dependable method to identify unsafe crossings, and unsafe movements, and reduce accidents by executing preventive management measures that exist in developing nations. Several countermeasures like providing speed breakers, speed humps, and speed tables can be provided to reduce ET and result in safer movements. An increase in the proportion of unsafe right turning over the years may also indicate the importance of implementing traffic control measures.

References 1. Allen B, Shin B, Cooper P (1978) Analysis of traffic conflict and collisions. Transp Res Rec No 677:67–74 2. Amundsen F, Hydén C (1977) Proceedings of first workshop on traffic conflicts. Institute of Transport Economics, Oslo, Norway 3. Gettman D, Pu L, Sayed T, Shelby S (2008) Surrogate safety assessment model and validation: Final report. Report No. FHWA-HRT-08–051. Federal Highway Administration, McLean. USA 4. Goyani J, Pawar N, Gore N, Jain M, Arkatkar S (2019) Investigation of traffic conflict at unsignalized intersection for reckoning crash probability under mix traffic condition. J East Asia Soc Transp Stud 13

608

C. N. M. Parvez et al.

5. Laureshyn A, Svensson A, Hydén C (2010) Evaluation of traffic safety, based on micro-level behavioral data: theoretical framework and first implementation. Accid Anal Prev 42:1637– 1646 6. Mohanty M, Panda B, Dey PP (2020) Quantification of surrogate safety measure to predict severity of road crashes at median openings. IATSSR-00257:7 7. MoRTH (2019) Road Accidents in India. Ministry of Road Transport & Highways, Transport Research Wing, New Delhi 8. Paul M, Ghosh I (2018) Speed-based proximal indicator for right-turn crashes at Unsignalized Intersections in India. J Transp Eng, Part A: Systems 144(6):04018024 9. Paul M (2019) Safety assessment at unsignalized intersections using post-encroachment time’s threshold—a sustainable solution for developing countries. In: Advances in transportation engineering. Springer, Singapore, pp 117–131 10. Perkins SR, Harris JL (1967) Criteria for traffic conflict characteristics. Report GMR 632. General Motors Corporation, Warren, Michigan 11. Reddy S, Chepuri A, Arkatkar S, Joshi G (2019) Developing proximal safety indicators for assessment of un-signalized intersection–a case study in Surat city. Transp Lett 12. Shekhar Babu S, Vedagiri P (2016) Proactive safety evaluation of a multilane unsignalized intersection using surrogate measures. Transp Lett. https://doi.org/10.1080/19427867.2016. 1230172 13. Shekhar Babu S, Vedagiri P (2017) Traffic conflict analysis of unsignalized intersection under mix traffic condition. Eur Transp/Transp Eur (66):10 14. Vedagiri P, Killi DV (2015) Traffic safety evaluation of uncontrolled intersections using surrogate safety measures under mixed traffic conditions. Transp Res Rec 2512(1):81–89

Ranking-Based Methodology for Prioritization of Critical Pedestrian Infrastructure in and Around the Market Area: A Case Study of Aminabad Market, Lucknow Haroon Rasheed Khan, Mokaddes Ali Ahmed, and Manish Dutta

Abstract Market areas are one of the busiest urban areas that experience heavy pedestrian footfall. However, studies on assessing existing pedestrian infrastructure are limited to market areas. This study addresses the gap by ranking the roads in and around the market area based on factors related to the road environment so that critical pedestrian infrastructures can be identified for improvement. Aminabad, a market area in Lucknow, Uttar Pradesh, has been taken as the case study area. As a renowned market centre and a focal point for tourism and cultural activities, many populations use walking as a transportation mode for their daily commute into the market area. The results show that the border roads of the market area scored an excellent rank in terms of safety and mobility due to the presence of proper pedestrian facilities compared to the inner streets. Based on the result, market roads have been divided into three categories: first, whose condition is good and does not need any repair; second, which requires a little care; last, which are critical and need immediate action. Keywords Pedestrian · Built environment · Walkability · AHP · TOPSIS · Market area

H. R. Khan (B) · M. A. Ahmed National Institute of Technology Silchar, Silchar, Assam 788010, India e-mail: [email protected] M. A. Ahmed e-mail: [email protected] M. Dutta Nirma University, Ahmedabad, Gujarat 382481, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_48

609

610

H. R. Khan et al.

1 Introduction As one of the world’s largest democracies, India is quickly advancing toward its goal of becoming one of the leading economic powers. As a result, urbanization has increased, and the number of people moving toward cities has also increased. Due to economic growth, vehicle ownership has also increased, leading to alarming traffic congestion on roads. However, to alleviate the heavy traffic congestion, roads are widened, and overpasses have been constructed, rather than trying to reduce the traffic on roads or to promote sustainable modes of transportation. To address the significant concerns of increased traffic, it is imperative to encourage walking as a preferred mode of transportation, at least for shorter trips, and public transit for longer distances. So, encouraging people to walk on foot might significantly cut down the number of trips being performed by automobiles in countries like India, where congestion is exceptionally high because of its dense population, shrinking spaces, and mixed land uses. Walking is one of the most basic and eco-friendly modes of transportation; everyone uses it for their daily commute. However, the ease of convenience and comfort offered by motorized vehicles has led people to depend on these modes even for shorter trips, contributing to traffic congestion. Furthermore, pedestrians are the most vulnerable road users and are more likely to be involved in an accident than other road users. Although a considerable proportion of trips are performed on foot in India, it needs proper pedestrian facilities. Traffic management of vehicular traffic is usually the primary focus of highway design in developing countries, with minimal emphasis on pedestrian facilities, which is a significant concern. Also, to cope with the rising volume of vehicular traffic, roadways are being widened at the cost of pedestrian facilities. As a result, the lack of sufficient pedestrian infrastructure has caused pedestrians to utilize the carriageway designated for automobiles, which not only interrupts traffic but also threatens their safety. This problem must be addressed by improving the design of pedestrian amenities so that more people will choose to walk instead of driving. Accordingly, this research paper examines the condition of the road environment in a market area regarding safety and mobility by assessing the various road environment factors or attributes that influence pedestrian movement. The objectives of this study are (1) to identify the gaps in the existing literature by performing a thorough assessment of the pedestrian walk environment and propose an appropriate methodology, (2) to identify the factors related to the road environment that influence the pedestrian movement in a market area, and (3) to prioritize the various roads of the market area based on the attributes of the existing road environment using AHP and TOPSIS. Section 2 presents a brief literature review on pedestrian walk environments and the various research work using the Likert scale for ranking. Section 3 explains the study’s overall methodology, including a brief overview of AHP and TOPSIS. Section 4 describes the study area and the data collection techniques. Section 5 presents the result and analysis starting with a preliminary analysis

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

611

and ranking various roads in the market area. Section 6 highlighted the findings, inferences, and future scopes that may provide necessary input to urban planners.

2 Literature Review A range of methodologies, including both qualitative and quantitative approaches, have been described in the literature. It is necessary to review the existing literature to better understand the research problem. Numerous studies have been conducted to prioritize transportation-related infrastructure projects [1, 2]. However, most of these investigations focused on road infrastructure development, maintenance, and rehabilitation [3, 4]. Only a few studies have been found in prioritizing transportation-related infrastructure [5, 6]. Prior research has shown that priority choices are driven mainly by several associated and competing decision factors, making it difficult for decision-making [3]. To address this, thorough consideration is needed to establish an appropriate set of regulating standards to overcome the abovementioned characteristics. Multicriteria decision-making (MCDM) approaches, such as analytic hierarchy process (AHP) [7], fuzzy AHP [2], and analytical network process [8], have been found to be systematic approaches [9, 10]. The prioritizing of attributes has been accomplished using several different multi-attribute decision-making (MADM) techniques. Such methods include grey relational analysis (GRA), analytical hierarchy process (AHP), elimination and choice expressing reality (ELECTRE), technique for order preference by similarity to ideal solution (TOPSIS), relative to an identified distribution integral transformation (RIDIT), etc. Among these, TOPSIS, one of the most extensively adopted MADM methods [11], is used in this study along with AHP. The TOPSIS method uses a derived scalar quantity for evaluation, giving an edge over other MADM approaches. Scalar value estimation is the ability to use a simple mathematical equation to compare the relative performance of the best and worst alternatives [12]. Researchers looked at how sidewalks, crosswalks at intersections, and midblock crosswalk affects the pedestrian level of service [13]. Few studies have developed a qualitative approach that considers pedestrian safety, security, continuity, comfort, and convenience. In some studies, pedestrian level of service was assessed using other factors such as traffic volume, surface quality, and obstructions. Few studies have focused on the safety considered by the adjacency to traffic flow and the level of segregation from the traffic. Most of the existing studies are based on user perception. Another similar kind of work has been found to assess the factors that would promote walking using people’s perception of identified factors [14]. Accordingly, a 9-pointer Likert scale questionnaire was constructed to evaluate user perception. Aesthetics and amenities, signage and street furniture, personal safety, and separation from traffic flow are some attributes included in the survey. Analysis and assessment of road segments were conducted using the analytical hierarchy approach [15]. Very little

612

H. R. Khan et al.

information was found in the existing literature that prioritized road environments for pedestrian movement using expert opinion-based data. Limited research work has been found on the pedestrian-built environment in the country. Assessment and prioritizing attributes based on expert opinion would assist in creating a pedestrian-friendly built environment. Road type, land use, average building frontage/setbacks, and road connectivity are the essential attributes found in the literature. The attributes are further divided into sub-attributes to determine the relative weights and effect of the attribute on the road environment, which are discussed in Sect. 4.

3 Methodology A detailed methodological framework for the study is shown below in Fig. 1. Fig. 1 Methodological framework

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

613

3.1 Theoretical Background of Analytical Hierarchy Process (AHP) In 1980, Saaty introduced the analytical hierarchy process (AHP), which has been widely used in most MCDM-related problems [16]. Educational institutions, engineering firms, government agencies, industrial sectors, and a variety of other sectors are using the AHP for decision-making [17]. Due to its simplicity, ease of use, and high versatility, it is widely applicable. The process of analytical hierarchy process is discussed in detail below. Step 1: Model development The hierarchy is established by dividing the issues into a series of interconnected decision points.

Step 2: Construction of pairwise comparison matrix among all the available attributes for that specific problem and assign the weights among the attributes using the fundamental scale given by Saaty.

614

H. R. Khan et al.

Step 3: Calculation of relative weights of all the attributes using the formula using Eq. (1) A1 A1 + A2 + A3 + ... + An A2 = A1 + A2 + A3 + ... + An .. . An = A1 + A2 + A3 + ... + An

W A1 =  W A2 .. . W An

(1)

Step 4: The consistency check, using the consistency index formula as shown in Eq. (2). Consistency I ndex(C I ) =

λm − n n−1

(2)

Step 5: The overall priorities were derived by aggregating the relative weights of various attributes as obtained from Step 3 to get the final decision of the problem, which is used as a rating for multiple decisions (or selections) to achieve the most general objective of the problem.

3.2 Theoretical Background of Technique of Order Preference Similarity to the Ideal Solution (TOPSIS) Hwang and Yoon introduced TOPSIS in 1981 to help select the best option based on a limited set of criteria. The typical TOPSIS technique attempts to choose alternatives with the longest distance from the negative ideal solution and the shortest distance from the positive ideal solution simultaneously [18]. TOPSIS is a wellknown MCDM approach that has received much attention from academicians and practitioners. The detailed process of evaluating TOPSIS is discussed below. Step 1: Objective determination and identification of pertinent evaluation attributes Step 2: Construction of normalized decision matrix as shown in Eq. (3) xi j r i j =  m j=1

(3) xi2j

Here, xi j and ri j are original and normalized scores of the decision matrix, respectively.

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

615

Step 3: Check for the relative importance weights of different attributes with respect to the objective, such that the summation of all the weights should be equal to one. n 

wj = 1

j=1

Step 4: Construction of weighted normalized decision matrix using formula as shown in Eq. (4). Vi j = w j ∗ ri j

(4)

Here, w j is the weight for j criterion. Step 5: Determination of positive ideal and negative ideal solutions using Eqs. (5) and (6).   A+ = Vi+j , ....., Vn+ Positi ve ideal solution

(5)

  Here, Vi+j = max(Vi j )|i ∈ I ; min(Vi j )|i ∈ I    A− = Vi−j , ....., Vn− N egative ideal solution

(6)

  Here, Vi−j = max(Vi j )|i ∈ I ; min(Vi j )|i ∈ I  Step 6: Calculation of separation measures for each alternative using Eqs. (7) & (8).

n

 2 + Sj =  (Vi j − V j+ ) Positi ve ideal solution

(7)

i=1



n

 2 − (Vi j − V j− ) N egati ve ideal solution Sj = 

(8)

i=1

Step 7: Evaluation of relative closeness to the ideal solution using Eq. (9). Ci∗ =

S −j S +j + S −j

(9)

Step 8: Based on the obtained value, sort the value of Ci∗ in decreasing order and rank it accordingly.

616

H. R. Khan et al.

3.3 Spearman’s Rank Correlation Coefficient In this research, two different ranking methods have been used to determine the ranking of various roads in the market area. So, deciding the final rank order derived from the specific ranking methods might be challenging. Therefore, a statistical test known as Spearman’s rank correlation coefficient (ρ) has been evaluated to determine the relationship between rank-order results.  6 di2 ρ =1− n(n 2 − 1)

(10)

Here, di = difference in the paired ranks. n = number of roads in the market area.

4 Study Area and Data Collection 4.1 Area of Study This research aims to ascertain the pedestrian-built environment towards the factors affecting their safety and mobility regarding the prevailing pedestrian amenities in Lucknow, the capital of Uttar Pradesh. As the capital city, Lucknow has an intriguing demographic mix. It not only attracts a large number of tourists but also generates a great deal of business. As a renowned market centre and a focal point for tourism and cultural activities, many populations use walking as a transportation mode for their daily commute into the market area. Despite the high volume of pedestrian travel, the city lacks adequate pedestrian infrastructure. Most of the existing pathways have been encroached on by hawkers or blocked by poles or vehicles parked on the sidewalks; this compels pedestrians to use the carriageway, which is already congested due to high vehicular traffic. An on-street survey based on existing attributes of the road environment was conducted in one of the central market areas of Lucknow (Aminabad) to ascertain the existing condition of the walking environment that affects the safety and mobility of pedestrians. Figure 2 shows the street view of the Aminabad market.

4.2 Data Collection After an intensive analysis of the relevant literature, a list of factors related to the road environment was compiled. From that list, a set of the most significant factors relevant to the typical Indian road context has been chosen. The data collection is

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

617

Fig. 2 Street view of Aminabad market

an auditor-based survey on a 5-pointer Likert scale on those attributes of the road environment. After reviewing the existing literature, this Likert scale was designed, but after collecting the field data, it was realized that most of the attributes fit in a 3-pointer Likert scale. So, we didn’t push it to the limits to fit our data into the 5-pointer Likert scale, and we opted for the scale as it is, either a 4-scale or 3-scale. So, based on the existing literature and experience gained from the reconnaissance survey, a list of attributes and sub-attributes was created, and the process of selecting the attributes and sub-attributes is called survey instrument design. Two auditors have been assigned to collect the primary data of the market area based on the predefined attributes of the existing road environment. After collecting data, the irrelevant attributes are removed in the first stage of data screening and sorted according to our

618

H. R. Khan et al.

objectives. Entropy Weight Method (EWM) has been used to evaluate each attribute’s weight related to the road environment.

5 Results and Analysis 5.1 Preliminary Analysis Preliminary data analysis has been presented in this section. The market area of Aminabad has been segmented into smaller parts by selecting the various roads, as shown in Fig. 2. Five significant roads surrounding the market area and ten minor roads inside the market area are selected for the study. The inner roads have been selected based on two parameters–(1) the most famous roads, and (2) one end of the road should be connected to any of the border roads of the market area except for Jute Wali Gali, which has been selected due to its high popularity. The area of Aminabad is 1.09 km2 , and the road lengths, along with the road markings of the various roads, are presented in Table 1. A customized Likert scale has been designed for the study area as per the various attributes of the road environment to assign the weights for sub-attributes. A table with attributes, sub-attributes, their respective scales, and the reason for assigning weights has been presented in Table 2. Table 1 Road lengths of the market area

Road no

Road name

Road length (m)

1

Subhash Marg

1680

2

Gurudwara Road

345

3

Latouche Road

1650

4

Nazirabad Road

575

5

Gangaprasad Marg

1110

6

Shri Ram Road

190

7

Kasai Bara Road

340

8

Arya Mandir Road

300

9

Niyamatulla Road

195

10

Jute Wali Gali

245

11

Ganeshganj Marg

1250

12

Lt Ved Tiwari Street

375

13

Khursheed Bagh Road

385

14

Dugawan Road

500

15

Nala Fatehganj

825

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

619

Table 2 Likert scale of different attributes Attributes

Sub-Attributes Likert-Scale Reason

Road type (As per IRC)

Arterial Road

4

Sub-arterial Road

3

Collector Street

2

Local Street

1

Residential

3

Mixed Land Use

2

Commercial

1

Wide (more than 9.0 m wide)

3

Moderate (6.0 m to 9.0 m wide)

2

Narrow (less than 6.0 m)

1

Land use (As per study area)

Average building frontage/ setbacks (As per national building code of India, 2011)

Road − connectivity (No. of roads connected)

Higher-order Road got a higher Likert Scale [due to better safety and mobility] (because of the wider carriageway; proper traffic lights, and signals; appropriate markings, etc.)

Higher the traffic areas, lesser the Likert scale (Due to poor mobility and safety)

Likert Scale is directly proportional to the Frontage length

−

It will be a numerical value, as the number of roads connected to a particular road

5.2 AHP-TOPSIS-Based Prioritization of Roads EWM has been used to evaluate the respective weights (W j ) of the attributes based on the inputs collected from the field survey using Eq. 11 and the weights are shown in Table 3 The result shows that land use got the highest weight among all the attributes as it is one of the significant parameters of road environment, whereas road connectivity has the least weightage. 1 − Ej j=1 (1 − E j )

W j = m

(11)

Table 3 Weights of attributes of road environment

Wj

Road type

Land use

Average building frontage/setbacks (m)

Road connectivity

0.2529

0.2622

0.2561

0.2288

620

H. R. Khan et al.

Here, E j = entropy measure. Based on the AHP and TOPSIS scores, the prioritization of roads has been evaluated as shown in Table 4. A larger value of AHP or TOPSIS scores for a specific road indicates that the road has been ranked higher based on existing road environment parameters. The result shows that the border roads got the higher rank due to the presence of desirable pedestrian facilities as compared to the inner streets, except Ganeshganj Marg, which brought the second rank due to the presence of a better road environment for pedestrians. The inner roads have got a lower rank due to the unavailability of the proper pedestrian facilities. Spearman’s rank correlation has been carried out to check whether the two rankings conform with each other. The correlation coefficient has been found to be 0.996 which is an indication of excellent conformity. These results lead to several key insights. Firstly, the minor roads inside the market area have received a lower rank due to the poor road environment for pedestrian movement. Secondly, the border roads have got a higher rank due to the availability of proper pedestrian amenities that will help pedestrians to roam in the market area easily and safely. Thirdly, Shri Ram Road, Kasai Bara Road, Arya Mandir Road, Niyamatulla Road, and Jute Wali Gali, these five roads that require immediate attention as they got a low order rank by both the ranking methods. These are the inner streets of the market area having less width and narrow building frontage (>6 m), comparable to the major roads. Moreover, the frontage area has been encroached on by shop owners or hawkers or electric poles, due to which building frontage/ setbacks are reduced, leading to traffic jams and safety concerns. Out of the four road environment attributes that are used in this study–road type, land use, and road connectivity cannot be changed. However, average building frontage/setbacks can be adjusted to provide a better road environment for pedestrian movement in the market area. Fourthly, those roads that rank between six and ten will also require minor interventions to enhance overall safety and mobility.

6 Conclusion and Future Scope The development of pedestrian infrastructure facilities is cost-intensive, like other urban infrastructure. In such cases, urban regions need a proposal for pedestrian infrastructure improvement based on existing road environment data and plan the infrastructure development in a phased manner. Therefore, a new methodological framework has been proposed in this research work which can be used as a tool to build a logical ranking order for market areas to prioritize their improvement needs. Existing road environment attributes are used in the proposed framework to characterize the walking environment of the Aminabad market in Uttar Pradesh. Based on the ranking done by AHP and TOPSIS, Shri Ram Road, Kasai Bara Road, Arya Mandir Road, Niyamatulla Road and Jute Wali Gali are the roads, which require immediate intervention to improve pedestrian safety and mobility.

Ranking-Based Methodology for Prioritization of Critical Pedestrian …

621

Table 4 Ranking of roads based on road environment characteristics Road no

Road

AHP score

TOPSIS score

AHP rank

TOPSIS rank

Remarks

3

Latouche Road

0.83

0.7072

1

1

11

Ganeshganj Marg

0.785

0.6549

2

2

5

Gangaprasad Marg

0.721

0.532

3

4

Good condition Does not require any intervention

1

Subhash Marg

0.712

0.6027

4

3

2

Gurudwara Road

0.673

0.4924

5

5

4

Nazirabad Road

0.587

0.3511

6

6

15

Nala Fatehganj

0.417

0.278

7

7

12

Lt Ved Tiwari 0.379 Street

0.1808

8

8

13

Khursheed Bagh Road

0.372

0.1598

9

9

14

Dugawan Road

0.368

0.1491

10

10

7

Kasai Bara Road

0.353

0.1051

11

11

8

Arya Mandir Road

0.345

0.0825

12

12

6

Shri Ram Road

0.342

0.071

13

13

10

Jute Wali Gali 0.327

0.0241

14

14

9

Niyamatulla Road

0

15

15

0.319

Fair condition Requires a little intervention

Critical condition Requires immediate intervention

This research is one of the unique efforts to prioritize roads in a typical Indian environment (market area) based on the existing road environment attributes and demonstrates an application of MCDM techniques like AHP and TOPSIS. Overall, these results might serve as basic suggestions for Indian planners and stakeholders. Although the conclusions are case-specific, the methods exhibited may be used in other cities with comparable road environment conditions. In this study, only road environment attributes have been considered to evaluate the safety and mobility of the market area, but some other attributes like factors related to road traffic can also be incorporated to get a clearer understanding. This work can be escalated by taking other factors related to pedestrian safety and mobility and then the overall market area can be ranked. Similarly, different market areas of the city can be ranked for

622

H. R. Khan et al.

comparison. The overall rank of a city may be obtained on the aggregation of the ranks of the different market areas.

References 1. Maitra B, Azmi M, Ibrahim SN (2002) Prioritization of road projects-a dis utility-based approach 2. Moazami D, Behbahani H, Muniandy R (2011) Pavement rehabilitation and maintenance prioritization of urban roads using fuzzy logic. Expert Syst Appl 38(10):12869–12879 3. Tsamboulas DA (2007) A tool for prioritizing multinational transport infrastructure investments. Transp Policy 14(1):11–26 4. Pal S, Maitra B, Sarkar JR (2016) An approach for prioritization of state highways and its application. Transp Dev Econ 2(2) 5. Berechman J, Paaswell RE (2005) Evaluation, prioritization and selection of transportation investment projects in New York City 6. Sadhukhan S, Banerjee UK, Maitra B (2015) Commuters’ perception towards transfer facility attributes in and around metro stations: Experience in Kolkata. J Urban Plan Dev 141(4) 7. Dalal J, Mohapatra PKJ, Mitra GC (2010) Prioritization of rural roads: AHP in group decision. Eng Constr Archit Manag 17(2):135–158 8. Cheng EWL, Li H, Yu L (2005) The analytic network process (ANP) approach to location selection: a shopping mall illustration 9. Sadiq R, Tesfamariam S (2009) Environmental decision-making under uncertainty using intuitionistic fuzzy analytic hierarchy process (IF-AHP). Stoch Env Res Risk Assess 23(1):75–91 10. Kabir G (2015) Selection of hazardous industrial waste transportation firm using extended VIKOR method under fuzzy environment 11. Yoon KP, Hwang C-L (1995) Multiple attribute decision making: an introduction. Sage Publications 12. Hung C.-C, Chen L-H (2009) A fuzzy TOPSIS decision making model with entropy weight under intuitionistic fuzzy environment, pp 2210–2213 13. Kadali BR, Vedagiri P (2016) Review of pedestrian level of service: Perspective in developing countries. Transp Res Rec 2581:37–47 14. Ariffin RNR, Zahari RK (2013) Perceptions of the urban walking environments. Procedia– Social Behav Sci 105:589–597 15. Zainol R, Ahmad F, Nordin NA, Aripin AWM (2014) Evaluation of users’ satisfaction on pedestrian facilities using pair-wise comparison approach. In: IOP Conference Series: Earth and Environmental Science, vol 18, no 1 16. Ho W (2008) Integrated analytic hierarchy process and its applications–A literature review. Eur J Oper Res 186(1):211–228 17. Vaidya OS, Kumar S (2006) Analytic hierarchy process: An overview of applications. Eur J Oper Res 169(1):1–29 18. Behzadian M, Khanmohammadi Otaghsara S, Yazdani M, Ignatius J (2012) A state-of the-art survey of TOPSIS applications. Expert Syst Appl 39(17):13051–13069

Machine Learning Categorization Algorithms for Traffic Conflict Ratings Pushkin Kachroo, Anamika Yadav, Ankit Kathuria, Shaurya Agarwal, and Mahmudul Islam

Abstract This paper provides a review of unsupervised and unsupervised categorization algorithms using safety surrogate data to predict the severity of traffic conflicts using processed traffic video data at conflict sites as well as human ratings of the conflicts. Keywords ML categorization · Safety surrogates · Traffic conflict

1 Introduction This paper presents a review of Machine Learning (ML) algorithms that attempt to solve the categorization problem. There are two types of algorithms for categorization: supervised and unsupervised. In the supervised algorithm, the input and output data is used to train the categorization algorithm to minimize the norm of the error difference of the data output from the model output for the same given input. In the unsupervised case, we use the data input and perform the categorization without the use of the output data. In general, the unsupervised algorithms are used when the output data is not available. Our study is on the topic of traffic safety where we attempt to categorize the severity of a traffic conflict in terms of categorical levels. In turn, the traffic conflicts are a surrogate safety measure for crashes. Hence, categorization of the traffic conflicts has a relationship to the probability of actual crashes and their severity [1, 2].

P. Kachroo (B) Department of Electrical and Computer Engineering, University of Nevada Las Vegas (UNLV), Las Vegas, NV, USA e-mail: [email protected] A. Yadav · A. Kathuria Department of Civil Engineering, IIT Jammu, Jagti, India S. Agarwal · M. Islam Department of Civil Engineering, University of Central Florida (UCF), Orlando, FL, USA © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_49

623

624

P. Kachroo et al.

Fig. 1 Safety surrogate data collection

In this paper, we use the various categorization algorithms for traffic conflict severity using the field data and human-derived categorization for supervised learning, and using direct field data for unsupervised learning. The data was collected in the field at various intersections of different types using fixed cameras as well as using drone videography overhead. The video data went through image processing to obtain vehicle trajectories, which were further processed for obtaining the variables used in the algorithms presented in this paper. Humantrained agents were used for categorizing the traffic conflicts for use in training the supervised algorithms. Figure 1 shows one of the locations from the data collection site.

1.1 Background Machine Learning (ML) has been studied and deployed in many applications for many decades starting from one of the early publications on the topic [3] and followed by the first learning rule in [4]. These lead to the development of neural networks. Modern ML also uses techniques from multivariate statistics for pattern recognition [5, 6]. Machine learning has been used in many applications in transportation for a while [7] and it is still an actively pursued area [8]. It is also being used in transportation safety area [9]. In this review paper, the authors show what work has been done in using ML for crash predictions.

1.2 Research Gap and Contribution Traffic accidents can qualify to be categorized as rare event phenomenon [10–12]. This can be seen by dividing the time variable into small increments such as in a few

Machine Learning Categorization Algorithms for Traffic Conflict Ratings

625

seconds or minutes and then looking at how many intervals involve accidents versus how many have no accidents. Since in any contiguous data collected on data, most of the data will have no accidents, it is very difficult to build estimation of accidents based on that. Hence, researchers have developed Safety Surrogate Measures (SSM) for accidents such as Time to Crash (TTC) and its various modifications, namely, time-exposed time to collision (TET) and time-integrated time to collision (TIT), post-encroachment time (PET), delta-V (.V), relative speeds, accelerations, etc. These SSMs have been used in the models for the prediction of traffic crashes as well as their severities [1, 2].

1.3 Problem Statement The current ML algorithms concentrate on crash predictions. In this current paper, we use ML techniques to connect some available SSMs to the ratings that have been given by human raters after being trained and watching the same videos that have been processed using software to perform video processing and then obtaining the SSMs automatically. Hence, the aim of this paper is to review the categorization ML algorithms and show how they can be used to mimic human rating method.

2 Algorithms The general framework of this paper is shown in Fig. 2.

Human Agent Processing Supervised −→

+

Traffic Video Data

− Image & Trajectory Processing

Un-Supervised −→

Human Categorization

Traffic Video Data

Image & Trajectory Processing

Fig. 2 ML cataegorization algorithms for ratings

ML Categorization

ML

ML

ML Categorization

626

P. Kachroo et al.

2.1 Supervised Algorithms In this paper, we have used three regression models that fall under the category of supervised algorithms, namely: (1) simple linear regression, (2) multiple linear regression, and (3) logistic regression. The data set we used for testing the models contains various variables such as vehicle type, time-exposed TTC (TETTC), timeintegrated TTC (TITTC), TTC Min (TTCM), speed of the vehicles, conflict type, and crash severity ratings. As we tried to predict the crash severity ratings of the incidents, we used the rating variable as the dependent variable for all the models throughout the manuscript. Simple linear regression tries to predict the quantitative response of the dependent variable (.Y ) based on a single independent variable (. X ). The equation for simple linear regression is given as .Y ≈ β0 + β1 X (1) Here, we fit the model using the TETTC and the models estimated the coefficients as .β0 = 0.34 and .β1 = 0.15. The residual standard error for this model is 0.485 and the . R 2 = 0.043. The regression line for simple linear regression is shown in Fig. 3, which tells us that with the increase of TETTC, the crash severity ratings also increase. Multiple linear regression uses at least two variables to predict the dependent variable. If we have .n distinct independent variables, then the equation for the dependent variable (.Y ) will be Y ≈ β0 + β1 X 1 + β2 X 2 + · · · + βn X n

(2)

0.0

0.2

0.4

Ratings

0.6

0.8

1.0

.

0

1

2

3 Time Exposed TTC

Fig. 3 Regression plot for ratings on time exposed TTC

4

5

Machine Learning Categorization Algorithms for Traffic Conflict Ratings

627

We used TETTC, TITTC, and the difference between the speed of the vehicles as independent variables to model the multiple linear regression algorithm. After fitting the models, we find the values for the coefficient as .β0 = 0.32, .β1 = −0.49, 2 .β2 = 0.145, and .β3 = −0.003. The residual value for the model is 0.45 and . R = 0.2. The correlation between the dependent and independent variables is given in Table 1. Logistic regression predicts the probability of an event occurring based on the independent variables. The probability of whether an event will occur or not can be modeled as . p(X ) = β0 + β1 X (3) For logistic regression, we used all the variables (vehicle type, vehicle speed, conflict types, TETTC, TITTC, and TTCM) to estimate the probability of the severity ratings of a crash. The prediction of this model is given in Table 2, which shows the model predicts the crash severity ratings at an accuracy of .81%. Support vector machine is a supervised learning framework that classifies the data using a hyperplane. We have used a linear hyperplane for the classification of the data. We used TETTC and TITTC variables to classify the crash severity ratings. First, we split the data into training and testing as a ratio of 7:3. The training data set contains 70% of the data used for the model’s training, and the rest 30% is used for validation. The SVM classification plot is shown in Fig. 4. The trained SVM model is tested on the testing data set. The result shows that the model predicts with an accuracy of 69%. The prediction value is given in Table 3. An artificial neural network (ANN) is an interconnection of nodes that can map the outputs with the inputs without any information on the relationship among the variables. ANN is typically made of one input layer, one output layer, and one or more hidden layers. In our case, we used vehicle types and speed differences in the input layer and the ratings in the output layer. We used one hidden layer with two

Table 1 Co-relation between dependent and independent variables for multiple linear regression TETTC TITTC Speed diff. Ratings TETTC TITTC Speed diff. Ratings

1

0.933 1

0.008 .−0.207

1

Table 2 Prediction made by the logistic regression model Predicted value Actual value 0 1

False 235 49

True 47 162

0.209 0.325 .−0.188 1

628

P. Kachroo et al.

Table 3 Prediction made by the SVM model Predicted value Actual value 0 1

0 76 8

1 37 26

SVM classification plot x

5

4

1

x

o o

o

o

o

3 TETTC

o x o o x

x

0

x o x

x x

o

x

o

o

o

o

oo o oo

0

o o x x xx o x o o x xo x x oo 1 x x o o x xo o x xxxx x o oo o x o ox x x xx x x o o o xo xxoxxx xxxx xxoooo oo ox xxxx xxxxx xx o o ooooo oo ooxo ooxo xxxxxxxxxxxxxx x xxo o o ooooo oooo ooooo oxxo xxxxxxxxxxxxxxxxxxxxx x x o o o o x o x x x x x o x o o x o xxxxxxxxxxx x oo o oxoxo o oxo xoxxxxxxxxxxxxxx x x xxx xxxxxxxxx xxx 0 xx o

o

o

x x

o

2

x

o

oo

o

5

10

15

TITTC

Fig. 4 Classifying the data using support vector machine Fig. 5 Structure of the artificial neural network

1

1

0.7 2

4 645

1.1

227

6 089 0.0

TType1

−1

.0 02 3

107

0.06

8 09

95

1.

Ratings

TType2 −6

.2

7 2.2354

38 91

9

31

20

1.0 123

1

1.

SpeedD

8 0.6

207

Error: 55.973144 Steps: 13801

Machine Learning Categorization Algorithms for Traffic Conflict Ratings Cluster plot

Optimal number of clusters Elbow Method

1_76

3500

1_279 1_70 1_137 1_82 1_141 1_203 1_219 1_273 1_167 0_301 0_322 1_121 1_224 1_282 1_260 0_242 1_111 0_220 1_163 1_233 1_284 1_298 0_211 1_277 0_191 1_209 1_264 1_289 1_78 1_299 0_148 1_130 1_270 0_216 0_362 1_124 1_22 1_63 1_162 1_235 1_297 1_112 0_327 1_252 1_125 1_166 1_101 1_156 1_272 1_11 1_56 0_229 1_140 0_309 0_182 1_173 0_4351_332 1_373 1_346 0_288 0_186 0_249 1_171 1_90 1_149 0_339 0_430 1_144 1_315 1_198 1_265 1_177 1_65 1_2711_269 0_471 1_110 0_352 1_278 1_222 1_160 0_367 1_85 1_199 0_155 0_470 1_188 0_214 0_387 0_180 1_138 1_307 1_328 1_134 0_333 0_371 0_302 0_158 0_338 1_48 1_323 1_221 0_461 1_254 1_8 0_92 1_350 1_39 0_213 0_404 0_295 0_334 0_255 0_116 1_55 0_107 1_276 1_318 1_131_196 1_170 1_58 0_394 0_86 1_354 0_305 0_294 1_204 1_109 1_250 0_432 0_376 0_466 0_401 0_428 0_172 1_3 1_201 0_427 1_290 1_100 1_245 0_479 1_97 1_324 1_87 1_26 0_390 0_206 0_359 0_261 0_424 1_81 1_21 1_61 1_16 1_293 0_459 0_64 1_83 0_317 1_19 0_425 0_393 0_345 0_436 0_412 1_187 1_168 0_395 1_98 0_750_95 0_358 1_291 0_453 0_357 0_491 1_157 0_341 1_194 0_372 1_193 0_319 0_306 0_464 1_28 0_152 1_251 0_450 0_321 1_89 0_356 0_405 1_106 0_310 0_205 1_108 1_14 1_37 1_74 1_41 1_73 0_286 0_434 0_399 0_455 0_478 0_238 1_218 1_231 0_215 0_474 0_383 0_258 0_361 1_388 0_419 0_175 1_159 0_320 1_145 0_176 0_304 1_5 1_51 0_442 0_379 0_292 1_12 1_40 1_146 0_316 1_117 0_353 1_7 1_139 0_263 0_377 0_246 0_489 0_351 0_382 0_248 0_460 0_300 1_337 0_184 1_88 0_381 1_118 0_485 1_363 0_407 0_465 0_396 0_431 0_326 1_36 1_256 0_456 1_161 0_422 1_2 0_413 1_114 1_482 1_200 0_384 0_287 0_420 0_375 0_408 0_406 1_66 0_391 1_15 1_62 0_417 0_490 0_313 0_364 0_398 0_454 1_239 0_348 0_403 1_17 1_6 1_35 0_447 1_237 1_771_135 1_1 1_49 0_448 0_120 1_71 0_400 0_473 0_481 1_151 0_266 1_127 0_457 0_230 0_202 1_25 0_429 0_113 0_330 0_440 0_280 0_487 0_262 0_179 0_409 0_126 0_150 1_10 1_169 0_228 0_411 0_443 1_18 0_178 0_343 0_259 0_380 0_438 0_104 0_369 0_189 1_27 0_426 1_183 0_129 1_38 1_212 1_80 1_192 0_223 0_416 0_463 0_303 1_102 0_50 0_227 0_142 0_232 1_47 0_449 0_153 0_147 1_128 0_329 0_236 0_1190_335 0_385 0_493 1_93 0_374 0_475 0_439 0_423 0_480 0_325 0_217 1_190 0_458 0_468 0_285 0_355 0_225 1_4 1_165 0_386 0_469 0_451 0_433 0_365 0_281 1_344 0_133 1_446 0_311 0_483 0_68 0_244 0_402 0_210 0_253 1_91 0_132 0_296 0_208 0_366 1_257 0_340 0_103 0_197 0_243 0_452 0_410 0_477 0_234 0_392 0_415 0_164 0_267 0_397 0_274 0_122 0_421 0_115 0_207 1_389 0_283 0_275 1_53 1_20 0_486 1_90_174 1_72 1_24 1_123 0_488 0_84 0_437 0_368 0_247 1_241 1_32 0_467 0_143 0_99 0_484 0_240 0_378 0_360 0_370 1_331 0_314 1_312 0_96 0_472 0_3080_268 1_541_34 1_43 0_4181_105 0_131 1_69 0_347 0_441 1_181 1_44 1_60 0_342 0_154 0_444 0_414 0_349 0_336 0_445 1_67 0_195 0_94 0_185 1_31 0_79 0_492 1_46 0_33 0_476 1_59 0_462 1_45 1_23 1_52 1_30 0_226

2

3000

2500

Dim2 (26.2%)

Total Within Sum of Square

629

0

2000 −2

1500

cluster

1_29

a

1

a

2

a

3

1_42

1_57 −4

1_136

1000 1

2

3

4

5 6 Number of clusters k

7

8

9

10

−2.5

0.0

2.5

5.0

7.5

Dim1 (29.1%)

(a) Estimating optimal number of clusters (b) Clustering the observation into three difusing elbow method ferent groups

Fig. 6 K-Means clustering on all the variables

nodes. After training the model, the weights are correctly tuned, shown in Fig. 5. This model can predict the crash severity ratings with an accuracy of 64%.

2.2 Unsupervised Algorithms K-Means Clustering is an unsupervised learning algorithm that tries to find a model which closely maps the inputs with the outputs without any levels. We used all the variables in the data set apart from the ratings to cluster all the observations. First, we used the elbow method to estimate the optimal number of clusters that encloses all the observations. From Fig. 6a, we can see that the optimal cluster number should be 3. Based on this, we cluster all the observations in three clusters, shown in Fig. 6b.

References 1. Arun A, Haque MM, Bhaskar A, Washington S, Sayed T (2021) A systematic mapping review of surrogate safety assessment using traffic conflict techniques. Accid Anal Prev 153:106016 2. Tarko A (2019) Measuring road safety with surrogate events. Elsevier 3. McCulloch WS, Pitts W (1943) A logical calculus of the ideas immanent in nervous activity. Bull Math Biophys 5(4):115–133 4. Hebb DO (1949) The organization of behavior: a neuropsychological theory. Science editions 5. Everitt B, Hothorn T (2011) An introduction to applied multivariate analysis with R. Springer Science & Business Media 6. Gareth J, Daniela W, Trevor H, Robert T (2013) An introduction to statistical learning: with applications in R. Spinger, Berlin 7. Arciszewski T, Khasnabis S, Hoda SK, Ziarko W (1994) Machine learning in transportation engineering: a feasibility study. Appl Artif Intell Int J 8(1):109–124

630

P. Kachroo et al.

8. Tizghadam A, Khazaei H, Moghaddam MH, Hassan Y (2019) Machine learning in transportation 9. Silva PB, Andrade M, Ferreira S (2020) Machine learning applied to road safety modeling: a systematic literature review. J Traffic Transp Eng (English edition) 7(6):775–790 10. Resnick SI (2008) Extreme values, regular variation, and point processes, vol 4. Springer Science & Business Media 11. De Haan L, Ferreira A, Ferreira A (2006) Extreme value theory: an introduction, vol 21. Springer, Berlin 12. Castillo E (2012) Extreme value theory in engineering. Elsevier

A Fuzzy Logic Approach on Pedestrian Crossing Behaviour at Unsignalized Intersection M. Manoj, Vivek R. Das , and Nitin Kumar

Abstract The pedestrian behaviours on roads affect their safety to a greater extent. Of many behaviours exhibited by pedestrians, crossing speeds, gender and gaps accepted and rejected to cross the road are of major concern. In this paper, the study is made on behaviour of pedestrians at unsignalized T intersection at Channasandra within the capital city of Karnataka, Bengaluru by generating a fuzzy model. The data were collected at the study location using a videographic survey for a duration of one hour during the peak hours and the above-mentioned behaviours were extracted and subjected to analysis. For modelling, gaps accepted and crossing speeds of pedestrians were given as inputs in MATLAB’s fuzzy logic toolbox by framing a set of rules, and after defuzzification the choice of pedestrians to cross the road will be obtained from the model. Keywords Pedestrian behaviours · Gaps accepted · Crossing speeds · Fuzzy logic · Unsignalized

1 Introduction In recent times there has been a drastic increase in the number of accidents on roads caused due to vehicle–pedestrian collisions, as per MoRTH “Road Accidents in India” report of 2020, accidents corresponding to pedestrians in the year 2020 stood at 57,763 and that of death toll was 23,483 in India alone. With the increase in M. Manoj (B) Department of Construction Technology & Highway Technology, Dayananda Sagar College of Engineering, Bengaluru, Karnataka 560078, India e-mail: [email protected] V. R. Das Department of Civil Engineering, Ramaiah Institute of Technology, Bengaluru, Karnataka 560054, India N. Kumar Department of Civil Engineering, Dayananda Sagar College of Engineering, Bengaluru, Karnataka 560078, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_50

631

632

M. Manoj et al.

population around the globe, the general trend followed is that people settle in the urban areas. Due to this the capacity of roads in urban areas have been decreased. There is also poor planning in terms of safety provisions for non-motorized road users. The lack of space available in completely developed cities is also one of the reasons. On the other hand, people are reluctant to follow basic rules and try to risk their lives by not using the provided pedestrian facilities. These problems have been a common sight in the capital city of Karnataka, Bengaluru. Thus, this study is intended to examine the behaviour of pedestrians at unsignalized intersections. For this purpose, a T intersection was selected in the area of Channsansdra, Bengaluru. Descriptive analysis and fuzzy model were developed from the data extracted in the above-mentioned intersection. The concept of fuzzy logic which was developed by Lofti Zadeh, analyses the input data using approximate and human language [1]. The final model gives the choice behaviour of pedestrians using the input data of gaps accepted and crossing speeds.

2 Literature Review Most of the previous studies mainly focus on pedestrian demographic characteristics (such as age and gender) and how these characteristics influence road crossing behaviour. Such studies have focused on detailed experiments to find out the effect of age on road crossing decisions with the effect of vehicle distance or speed of vehicle [2, 3]. Khan et al. [4] studied pedestrian behaviours for three different activities in the city of Karachi, Pakistan. This paper also observes the effect of sidewalks and street encroachments on pedestrians. Statistical Analysis was done using the Fisher exact test with a 95% confidence level. The results were given on all three activities of pedestrians considered namely, Road crossing behaviour, Street pedestrian behaviour, Sidewalk pedestrian behaviour, and finally on encroachments. Dıaz [5] study gives the theory of pedestrian behaviour and their intentions to violate the rules for which data was obtained from self-rating methods from the sample of pedestrians and the evaluated using statistical analysis method EQS. Lapses, errors, and violations of different gender and age groups were reported. Chai et al. [6] used fuzzy logic to examine differences in cognition for age and gender traits at jaywalks and signalized crosswalks by observing pedestrian movements captured by videography survey. Fuzzy sets and rules were defined to model the decision–making process affected by signal timing, pedestrian group, upcoming vehicles, and ongoing pedestrians. Further, improvements can be made to refine and extend the applications of the fuzzy logic method. This paper also suggests to conduct Questionnaires surveys to further understand the behaviour, perception, and attitude of road users. Dutta and Vasudevan [7] studied the effect of the waiting time of pedestrians on the rolling gap acceptance at unsignalized intersections in heterogeneous urban traffic environments in six sites in Kanpur, India. In India, pedestrian guidelines are mentioned in IRC—103 [8]. The analysis was done using the survival analysis approach. The result showed the various factors that affect the rolling gap acceptance by pedestrians.

A Fuzzy Logic Approach on Pedestrian Crossing Behaviour …

633

Fig. 1 Map of selected intersection (Google image dated 15.06.2022)

3 Methodology 3.1 Site Location An unsignalized intersection was selected in Channasandra area within the city of Bengaluru, Karnataka. This location is comprised of 3 arms meeting together forming a T-junction. This intersection was selected due to more pedestrian footfall observed during the reconnaissance survey because of the presence of educational institutes and commercial activities near the intersection. Figure 1 shows the map of the location selected. Approach A connects to Mysuru Road, approach B connects to Uttarahalli, and approach C connects to the interior of RR Nagar.

3.2 Data Collection Data was collected from the selected site location using a videographic survey with the support of a DSLR camera. The survey was conducted for a duration of one hour during peak hours, in this case the peak hour was 9 am to 10 am. The camera was placed on the terrace of a high-raised building at the intersection in such a way that the view from the camera covered the entire area of the intersection road area. Figure 2 shows the setup of camera at study area.

634

M. Manoj et al.

Fig. 2 Camera setup at Channasandra junction

Table 1 Details of Channasandra junction

Pedestrian flow

210

Vehicle count

3316

Crosswalk marking

Yes

Presence of sign boards

No

Presence of median

No

3.3 Data Extraction The data such as pedestrian volume count, crossing speeds of pedestrians, and gaps accepted and rejected were extracted from the survey conducted through videographic survey based on the genders. Also, the few components in the vicinity of the study intersection, such as crosswalk markings and presence of sign boards, were noted. Table 1 shows a few details of the intersection. Male and female pedestrians percentage was 62% and 38% respectively of the total pedestrian volume count of 210. The average speed was found to be 1.34 m/s and 1.13 m/s for male and female pedestrians, respectively.

3.4 Descriptive Statistics The quantitative summary of statistics obtained from the easyfit software is represented in this section. The sample data were subjected to descriptive statistics analysis for two inputs namely gaps accepted and crossing speeds of pedestrians. The mean values from the sample data are shown for minor and major crossing along

A Fuzzy Logic Approach on Pedestrian Crossing Behaviour …

635

with the deviation of data from its mean values. Also, the value of skewness and excess kurtosis for gap and speed data is mentioned in the following Tables 2 and 4, respectively. Minimum, median, and maximum values along with some other percentiles of the sample data of gaps accepted and crossing speeds are tabulated in Tables 3 and 5. From this, the average crossing speeds are 1.32 m/s and 1.08 m/s for males and females, respectively. The minimum time gaps are accepted by males when compared to females. The average time gap accepted is 5.66 s and 5.44 s for males and females, respectively (Table 4). Table 2 Descriptive statistics of gap Statistic Range

Minor crossing

Major crossing

Male

Female

16.73

7.99

Male

Female

11.29

11.05

Mean

5.52

5.28

Variance

9.28

3.31

10.011

5.9691

7.41

5.68

Std. deviation

3.05

1.82

3.164

2.72

Coef. of variation

0.55

0.34

0.53007

0.48

Std. error

0.37

0.30

0.55079

0.52

Skewness

2.04

0.45

0.87648

1.26

Excess kurtosis

5.64

−0.02

−0.05431

1.41

Table 3 Percentile statistics of gap Percentile

Minor crossing Male

Major crossing Female

Male

Female

Min.

1.39

1.92

2.36

2.6

5%

2.10

2.67

2.40

2.72

10%

2.44

2.91

2.57

2.89

25% (Q1)

3.55

3.95

3.27

3.77

50% (Median)

5.13

5.22

5.56

4.49

75% (Q3)

6.45

6.76

8.26

6.73

90%

9.05

7.83

10.80

9.91

95%

12.38

8.90

13.26

12.35

Max.

18.12

9.91

13.65

13.65

636

M. Manoj et al.

Table 4 Descriptive statistics of speed Statistic

Minor crossing Male

Major crossing Female

Male

Female

Range

2.59

1.11

1.11

1.22

Mean

1.12

1.57

1.57

1.36

Variance

0.12

0.13

0.13

0.17

Std. deviation

0.34

0.36

0.36

0.41

Coef. of variation

0.31

0.23

0.23

0.31

Std. error

0.04

0.06

0.06

0.07

Skewness Excess kurtosis

3.07

0.19

0.19

0.05

15.84

−1.58

−1.58

−1.41

Table 5 Percentile statistics of speed Percentile

Minor crossing

Major crossing

Male

Female

Male

Female

Min.

0.64

0.55

1.09

0.76

5%

0.78

0.65

1.11

0.76

10%

0.81

0.75

1.15

0.80

25% (Q1)

0.90

0.81

1.25

0.93

50% (Median)

1.08

0.9

1.56

1.39

75% (Q3)

1.24

1.01

1.91

1.73

90%

1.43

1.16

2.04

1.96

95%

1.59

1.27

2.12

1.97

Max.

3.23

1.29

2.20

1.98

3.5 Distribution Data The data were subjected to distribution fittings in the easyfit software for both inputs (Gap Accepted and Crossing Speed) that are planned to be used as inputs in the MATLAB fuzzy logic. For the gaps accepted data there were four different distributions of which Dagum (3P), Gen. Extreme Value, and Lognormal (3P) have similar curves with varying equation parameters and changes in skewness and kurtosis. The distribution data of the gaps accepted is shown in Table 6. Similarly, for the data regarding crossing speeds of pedestrians, there were three different distributions with Johnson SB distribution repeating itself for both male and female pedestrians in the major crossing section. All the three distributions are smooth curves similar to normal distribution curves with varying equation parameters and changes in skewness and kurtosis. Cauchy is a leptokurtic distribution. The distribution data of speed is shown in Table 7.

A Fuzzy Logic Approach on Pedestrian Crossing Behaviour …

637

Table 6 Distribution data of gap Arm type

Minor crossing

Major crossing

Gender

Distribution

Kolmogorov Smirnov Statistic

Rank

Parameter

Male

Dagum (4P)

0.08462

1

k= 0.55762 α = 3.4679

β= 4.8467 γ = 1.1466

Female

Gen. extreme value

0.07561

1

k = −0.13449 σ = 1.666 μ = 4.5142

Male

Gen. Pareto

0.09544

1

k = −0.24169 σ = 4.9208 μ = 2.0061

Female

Lognormal (3P)

0.12465

1

σ = 0.7852 μ = 0.99376 γ = 2.0893

Table 7 Distribution data of speed Arm type

Gender

Distribution

Dagum (4P)

Kolmogorov Smirnov Statistic

Rank

0.0565

1

Parameter

κ = 0.51445 a β = 0.60585 γ = 4.201 = 0.592

Minor crossing

Male Female

Cauchy

0.09432

1

σ = 0.07323 μ = 0.89102

Major crossing

Male

Johnson SB

0.10274

1

γ = 0.13447 d λ = 0.95338 ξ = 0.3052 = 1.1403

Female

Johnson SB

0.07474

1

γ = 0.03976 d λ = 1.2862 ξ = 0.46971 = 0.72888

3.6 Fuzzy Logic Model Development Fuzzy logic is an easy yet effective tool that suits well for the analysis of data that involves randomness and vagueness. In this paper, pedestrian choice behaviour for crossing the road is modelled in MATLAB’S fuzzy logic with two variables, gaps accepted and speed of crossing as input parameters. Gaussian distribution had a very similar match with the distribution curves obtained from the easyfit software, tabulated and explained in earlier sections. Thus, for both inputs, Gaussian membership function was defined and a triangular membership function was defined with No (decision is not crossing) and Yes (decided to cross) as output. A set of rules were framed as tabulated in Table 8 against which fuzzy logic analysis for the data was done. The results were obtained in the form of surfaces and rules and are shown for minor and major crossings for both genders from Figs. 3, 4, 5 and 6.

638 Table 8 Rules for evaluation

Fig. 3 Minor crossing—male

Fig. 4 Minor crossing—female

Fig. 5 Major crossing—male

M. Manoj et al.

Input 1 gap

Input 2 speed

Output choice

Low

Low

No

Low

Medium

No

Low

High

No

Medium

Low

No

Medium

Medium

Yes

Medium

High

Yes

High

Low

No

High

Medium

Yes

High

High

Yes

A Fuzzy Logic Approach on Pedestrian Crossing Behaviour …

639

Fig. 6 Major crossing—female

4 Conclusions • The average time gap accepted by pedestrians in seconds was 5.66 s and 5.44 s for males and females, respectively. • The average crossing speed of pedestrians was found to be 1.32 m/s and 1.08 m/ s for males and females, respectively. Thus, it is clear that the walking speed of males is found to be faster compared to that of females. • The distributions for the gap accepted data as obtained from easyfit were Dagum (4P), Gen.Extreme Value for minor crossing corresponding to males and females, respectively, and Gen Pareto and Lognormal (3P) for major crossing corresponding to males and females, respectively. • The distributions for the crossing speed data as obtained from easyfit were Dagum (4P), Cauchy for minor crossing corresponding to males and females, respectively, and Johnson SB distribution for both males and females in major crossings. • Four models were developed each for minor crossing and minor crossings for two genders. These models help in the study of the choice behaviour of pedestrians at unsignalized intersections.

References 1. Zadeh LA (1965) Fuzzy Sets. Inf Control 8:338–353 2. Lobjois R, Cavallo V (2007) Age-related differences in street-crossing decisions: the effects of vehicle speed and time constraints on gap selection in an estimation task. Accid Anal Prev 39(5):934–943 3. Oxley J, Fildes B, Ihsen E, Charlton J, Day R (1997) Differences in traffic judgments between young and old adult pedestrians. Accid Anal Prev 29(6):839–847 4. Khan FM, Jawaid M, Chotani H, Luby S (1998) Pedestrian environment and behavior in Karachi, Pakistan. Accid Anal Prev 31(1999):335–339 5. Dıaz EM (2002) Theory of planned behavior and pedestrians’ intentions to violate traffic regulations. Transp Res Part F 5:169–175 6. Chai C, Shi X, Wong YD, Er MJ, Gwee ETM (2016) Fuzzy logic—based observation and evaluation of pedestrian’s behavioural patterns by age and gender. Transp Res Part F: Traffic Psychol Behav 40:104–118

640

M. Manoj et al.

7. Dutta B, Vasudevan V (2017) Study on pedestrian risk exposure at unsignalized intersection in a country with extreme vehicle heterogeneity and poor lane discipline. Transp Res Rec: J Transp Res Board 2634:69–77 8. IRC—103—2012: Guidelines for pedestrian facilities

A Review on Surrogate Safety Measures Using Extreme Value Theory Dungar Singh , Pritikana Das , and Indrajit Ghosh

Abstract Over the past few decades, the observed accident data has served as the primary source for road safety analyses. To avoid the shortcomings of crash data, traffic conflict techniques have been promoted as an alternative approach to analyze safety from a wider perspective than relying solely on crash data. Still, the application of surrogate safety measures, and the validity of conflict techniques is a great concern. The Extreme Value Theory (EVT) offers Peak Over Threshold (POT) and Block Maxima (BM) approaches to provide a robust modeling framework for relating surrogate safety measures (SSMs) to crash frequency. This study is a comprehensive review of the development and use of EVT in combination with surrogate measures for safety evaluation. A number of formulated research questions on the use of EVT for SSM analysis are also identified and discussed in a precise manner. The key finding of the study demonstrates that bivariate extreme value modeling approaches with a combination of surrogate measures, time to collision (TTC), and Post Encroachment Time (PET), are more useful to estimate safety than uni-variate extreme value modeling approaches with individual safety measures. Finally, the authors identified several research gaps to assist researchers and practitioners with recommendations for potentially useful avenues of future research. Keywords Surrogate safety measures · Extreme value theory · Peak over threshold · Block maxima

D. Singh (B) · P. Das Maulana Azad National Institute of Technology Bhopal, Bhopal 462003, India e-mail: [email protected] I. Ghosh Indian Institute of Technology, Roorkee 247667, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_51

641

642

D. Singh et al.

1 Introduction Road safety is a matter of great concern in developed countries with lane-based traffic conditions and developing countries with non-lane traffic conditions, where rapid growth of motorized vehicles is coupled with inefficient roadway infrastructure to control outcomes. Annually, approximately 1.3 million people die, and around more than 50 million suffer non-fatal injuries globally, World Health Organization [19]. Numerous factors associated with such a scenario are mainly inefficient driving behavior, poor road infrastructure, and lack of management. Over the years, in traditional road safety analysis, several approaches have been proposed, such as beforeafter studies, Elvik [6]; identification of black spot programs, Ahmed et al. [1]; statistical modeling and road safety audits, Meuleners et al. [12]. These methods primarily rely on past crash data, statistical modeling, and skilled and experienced field observation. This type of study is known as a reactive technique, this begins based on existing crash data and focuses on identification of the high-risk area, which implies that a significant amount of accidents are to be recorded to study road safety. A reactive approach has significant limitations, such as poor data quality and a lack of accident data. According to research by several previous researchers [2, 15, 18], in many cases, accidents that do not result in serious injury are not reported from the site. As a result of the aforementioned issues, it is necessary to offer an alternative safety evaluation approach. Safety evaluation using traffic conflict technique (TCT) applying different SSMs is another alternative approach to the crash-based events method. It is a safer approach for resolving all the above-described difficulties. Initially, General Motors Corporation Laboratories adopted the TCT Perkins and Harris (1967) to characterize the interaction between two road users in traffic collision conditions and estimate the probability of a crash. traffic conflict analysis is a proactive approach to safety evaluation that does not solely rely on actual crash data. However, with the additional implementation of an image processing system and sensor-based technology for video-graphic data extraction and driver behavior characteristic data using a driving simulator and naturalistic driving study. The validity of SSMs is typically accessed by correlation with actual crash data frequency, typically evaluated by regression analysis models [11, 20, 21]. Furthermore, some non-crash-based and more advanced approaches have been suggested by researchers in the evaluation of road safety, such as automated road safety analysis, extreme value theory, and the causal model approach [16, 17]. The extreme value theory offers two approaches, Block Maxima (BM) or Peak Over Threshold (POT), to examine the suitable threshold for surrogate safety measures. Application of BM and POT approaches is becoming increasingly important in conflict techniques approach in recent years. Firstly, Songchitruksa and Tarko [17] developed EVT based method to analyze road safety as an alternative to the classical regression analysis without relying solely on crash data. Some of the existing study applications for extreme value theory have been discussed nicely in Tarko,

A Review on Surrogate Safety Measures Using Extreme Value Theory

643

2018. This study provides a comprehensive review of the recent development of the extreme value theory. The study objective is as follows: 1. To understand the state of the art of EVT and its applications in conflict techniques; 2. To discuss the current problems on the application of EVT in the road safety analysis; 3. To identify research gaps for promising future research directions of SSM and their applications in EVT safety. The organization of this article is as follows: The first step in identifying relevant studies is to discuss the research aims and questions. The second step is to use a search technique, inclusion and exclusion criteria, and gather relevant literature. In order to identify limitations and research gaps, the current literature was further segmented into different and thoroughly discussed. Finally, the study concluded with a review of the results and suggestions for further research.

2 Methodology In order to provide a thorough summary of the current research-related evidence, a systematic literature review (SLR) is frequently conducted. It is a systematic procedure that can be used to synthesize earlier research done by academics, researchers, and practitioners. The objective of SLR is to reduce occurrence bias in a thorough literature search across multiple databases and provide an answer to a predetermined, focused research question [13]. This study follows “The Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols” (PRISMA) framework for analyzing previous studies.

2.1 Search Strategy For identification of relevant literature, the following scientific databases were searched; Science Direct, Scopus, and Google Scholar. The author uses different keywords, with their combinations also used to ensure comprehensiveness. These are as follows: “conflict technique”, “surrogate measure”, “Extreme value theory”, and “safety–critical event”. The snowballing approach was immediately applied after a first screening to locate additional pertinent missing papers. As mentioned, this methodological approach is designed to address all predefined research objectives. The total search strategy found 452 relevant studies for further analysis. The data extraction process was completed using SLR inclusion and exclusion criteria. Further relevant studies were exported into an Excel spreadsheet, and a two-step procedure was used to include pertinent systematic reviews:

644

D. Singh et al.

Table 1 Eligibility criteria for selecting relevant literature Sn. No

Basic criteria

Decision

1

Research presented and published in conference proceedings and peer-reviewed journals

Inclusion

2

Research defines and elaborates on key concepts, surrogate safety measures (SSM), extreme value theory (EVT)

Inclusion

3

Research is establishing the relationship between conflict and crashes

Inclusion

4

Inaccessible articles and review papers

Exclusion

5

An article that is duplicated among database search results

Exclusion

6

The study of work that is not based on original research

Exclusion

7

Studies based on surrogate safety in medical science

Exclusion

6

Grey literature and dissertations

Exclusion

1. All papers that were initially retrieved had their titles and abstracts independently reviewed by two reviewers. Discussion was used to try to reach an agreement in case of disagreement. 2. Two impartial reviewers carried out full-text screening on a few chosen systematic reviews. When there were disagreement, discussion was used to try to reach an agreement. 3. Which extreme value theory approach Block Maxima or Peak Over Threshold, can estimate the probability of conflict efficiently?

2.2 Eligibility Criteria The application of eligibility criteria guarantees ensures that the literature used in the systematic review is relevant to the study, producing more accurate, objective, and significant results. However, a systematic selection procedure that applies the eligibility requirements criteria to the search results in a way that minimizes the possibility of selection bias would be beneficial. Several eligibility criteria were applied in order to find the pertinent literature, as shown in Table 1.

2.3 Quality Assessment The quality evaluation has been done in multipart and corresponds to various objective studies. The following criteria were used to assess the quality assessment of the systematic literature review. 1. Are the review’s inclusion and exclusion criteria sufficiently described?

A Review on Surrogate Safety Measures Using Extreme Value Theory

645

2. Is it likely that the literature search included all relevant studies on surrogate safety assessment? 3. Did the chosen publication use blind reviewers to rate the study’s validity and quality? 4. Was the type of extreme value theory mentioned in the literature described adequately?

2.4 Study Selection The authors found 450 records in total using three online databases: Science Direct (80), Scopus (62), and Google Scholar (308). Apart from these database sources, the authors included two studies from other known sources. Initially, the duplicate types of literature in the database were excluded, and 350 articles remained, after which 180 were excluded during the title and abstract screening process. Furthermore, an eligibility assessment was carried out on potentially 213 full-length texts, in which 130 articles were excluded from the study because their content was irrelevant or the application of extreme value theory was not well defined. Finally, 10 full-length articles were included in the narrative synthesis. Time restrictions were not taken into account in this study. Studies must be written in English and published in peerreviewed journals and conference proceedings, including those that describe the application of EVT in surrogate safety assessment. However, the systematic literature did not include “Dissertations” or “Grey Literature”. The characteristics of the final selected studies are summarized in Table 2, and the literature mapping is described in Fig. 1.

3 Discussion 3.1 Extreme Value Theory Extreme value theory (EVT) is a branch of statistics dealing with the stochastic behavior of extreme events and deviations from the median of probability distributions. The three extreme value distributions (Gumbel, Frechet, and Weibull) can each be used to explain the behavior of maxima for a single distribution, Fisher and Tippett [8]. The key feature of the extreme value theory is to predict rare events that are larger and less frequent than previously observed events [5]. Normally, this extreme behavior is rare and unobservable in the data collection period. Extreme value theory in transportation engineering was initially applied by Hyde and Wright [10] to estimate road traffic capacity. Henceforth et al. (2006) estimate crash probability based on specific crash proximity measures. Extremes sampling methods are frequently a source of serious concern. There are typically two types of extremes from various sample approaches.

646

D. Singh et al.

Table 2 Summary of the literature on extreme value theory Sn No

Author name, year

Country

Road geometry

SSM measures

Type of conflict

Modeling techniques

1

Zheng et al. [20]

China

Freeway

PET ≥ 0 s

LC

BM, POT

2

Farah and Azevedo’s [7]

USA

RH

TTC ≥ 0

CC

BM, POT

3

Zheng et al. [21]

China

Freeway

PET ≤ 0 s

LC

POT

4

Borsos et al. [3] Belarus

SI

TTC ≤ 0 s T2 ≤ −2 s

CC

BM, POT, KS

5

Songchitruksa and Tarko [17]

USA

SI

PET ≤ 0 s

CC

BM

6

Chauhan et al. [4]

India

SI

TTC < 2 s DRAC > 3.4 m/ s2

RE

POT

7

Goyani et al. [9] India

UI

PET ≤ 0 s

CC

BM, KS

8

Zheng and Sayed [23]

SI

TTC ≥ 0 PET ≥ 0

CC

BM, POT, PC, KRC

9

Jonasson and Sweden Rootzén (2014)

Urban Road

TTC ≤ 0 s

RE

BM

10

Zheng et al. [22]

Highway

LPM > 1

LC

BBM

Canada

Canada

Note SI—Signalized Intersections, UI—Unsignalized Intersection, CS—conflicting speed of vehicle, MADR—Maximum Available Deceleration Rate, KS—Kolmogorov–Smirnov Test, RH— Rural Highway, LC—Lane change, RE—Rear End, CC—Crossing Conflict, UBM—Uni-variate Block Maxima, BBM—Bivariate Block Maxima, UPOT—Uni-variate Peak Over Threshold, BPOT—Bivariate Peak Over Threshold, PC—Pearson’s Correlation, KRC—Kendall’s Rank Correlation, SRC—Spearman Rank Coefficient, Light-GBM—light gradient-boosting machine, THW— end of the passing maneuver, HC—Head on Collision, PCC—Pearson correlation coefficient, AIC—Akaike Information Criterion, RL—Reinforcement Learning

3.1.1

BM Approach

Block Maxima (BM) approach observations are grouped into fixed intervals over time and space, a block of minima or maxima is treated as extremes. This shows the extreme values occurring at maximum over a period of time, such as an hour or a year. Whereas extreme event samples follow a Generalized Extreme Value (GEV) distribution, the samples are distributed in different blocks in a certain time interval, considering the largest and smallest threshold in each sample block. As BM approaches, R-largest order statistics are typically preferred as the majority of blocks have enough observations. Block maxima approach mainly focuses on the behavior of independent random observation distributions.

A Review on Surrogate Safety Measures Using Extreme Value Theory

647

Fig. 1 Flow chart of the inclusion articles in the review Source Liberati et al. (2009)

Mn = max{X1 , X2 , . . . Xn } where X1 , X2 ,…Xn is, Xn is a series of independent random variables with the same distribution function F(x) = Pr(Xi ≤ x). Mn is the process maximum over n time units of observation. When n → ∞, the function Mn will converge to GEV distribution [20, 21]. The GEV function is shown below in Eq. 1. −1/   (x − μ) F(x) = ex p{−1 1 + ξ ξ ]} σ

(1)

The three parameters are defined as location parameter (μ) − ∞ < μ < ∞, scale parameter (σ) σ > 0, and shape parameter (ξ) − ∞ < ξ < ∞. The distribution of the function is defined as per the value of shape parameter, ξ, Frechet distribution ξ > 0, Weibull distribution ξ < 0, and Gumbel distribution ξ = 0.

3.1.2

POT Approach

Pickands [14] treats the Peak Hour Threshold (POT) as an extreme value if the related measurement exceeds the optimal predetermined threshold, where random observations {X1, X2, …, Xn} are identically and independently distributed within the Generalized Pareto Distributions (GPD). For a suitable threshold, the GPD is as follows:

648

D. Singh et al.

ξ G(x; μ; σ; ξ) = [1 − (1 + (x − μ) − 1/ξ]} σ

(2)

where X − u = exceeds the threshold conditional on X > u, σ > 0 = scale parameter, − ∞< ξ < ∞ = the shape parameter.

4 Literature Review Summary The extreme value theory was developed in the early twentieth century by Fisher and Tippett [8]. The authors distributed the limiting frequency of extreme values across larger and smaller sample sizes. The concept of EVT provides the possibility of estimating the likelihood of extreme events for a relatively short period of observation. EVT is a method for analyzing safety quickly that is similar to SSMs. The approach produces a group of models that make it possible to extrapolate from frequently occurring events, like traffic conflicts, to less frequently occurring events like crashes [23]. In addition, EVT provides a single dimension to assess the severity of surrogate events and identifies actual crashes [17, 20]. EVT use in analyzing traffic safety has grown significantly in recent years. Initially, Songchitruksa and Tarko [17] proposed the EVT (Block maxima) method to evaluate the safety of right-turning movements at 18 signalized intersections for an 8-h duration by SSM indicator post encroachment time. The proposed method evaluation findings showed a correlation between model estimates and crash data. Henceforth, analyzing the safety of freeway, authors, Zheng et al. [20] used two approaches, BM and POT, to evaluate lane change maneuver behavior at several freeway spans. Studies done by Zheng et al. [20] and Borsos et al. [3] found that the POT approach performs better than the BM for short interval data in terms of data estimate accuracy and reliability. On the contrary, for long interval data, the BM approach yielded better accuracy than the POT approach, Farah and Azevedo [7]. Moreover, analyzing the lane change behavior, Zheng et al. [21] proposed a shifted Gamma-GPD model and parameters such as threshold μ and shifted value δ to estimate crashes. They compared the Bayesian approach to the conventional maximum likelihood estimation approach. However, the application of naturalistic driving in estimating near-crash safety continuum, the study by Jonasson and Rootzén (2014), used the bivariate BM approach, safety measures TTC, in combination with a number of explanatory variables (max speed, min distance, right lane marking, etc.) to estimate a crash probability. Farah and Azevedo [7] used the BM and POT approaches to evaluate head-on collisions during passing maneuvers on two-lane rural highways. For freeway safety, PET and length proportion of merging (LPM) were used by Zheng et al. [22] to categorize the seriousness of events in merging areas. The authors proposed combining a bivariate extreme value model with a bivariate GPD. The findings demonstrated that the estimates of crashes from the bivariate model were

A Review on Surrogate Safety Measures Using Extreme Value Theory

649

more precise and accurate than those from traditional uni-variate models. Additionally, Zheng and Sayed [23] compared bivariate (BGP and BGEV) and uni-variate (UGP & UGEV) extreme value models and found that conflict, estimated by bivariate extreme value model (GPD), performs better than bivariate GEV and conventional uni-variate extreme value models. Borsos et al. [3] used TTCmin and T2 and showed that the conflict estimated by TTCmin is more accurate, whereas SSM indicator T2 overestimated the precise crash probability. Zheng and Sayed [23] observed that bivariate extreme value GPD models outperform traditional uni-variate extreme value models for safety analysis. Moreover, for carrying out safety analysis in mixed traffic conditions, Chauhan et al. [4] used the POT approach to determine the threshold for conflict probability. They found the probability of conflict TTC threshold less than 2 s and DRAC threshold greater than 3.4 m/s2 . Furthermore, Zheng et al. [20] adopted a PET threshold of less than 1 s for defining the likelihood of crash, whereas Goyani et al. [9] selected both lower and upper limits (−6, 6) and used the PET threshold level to estimate the critical conflict (−1, 1).

5 Conclusions Traffic conflict techniques have been widely used in recent years to analyze the safety of different traffic facilities (Intersections, Mid-Blocks, and Highways) in real time and provide remedial measures before and after safety analysis. This study reviewed several existing studies on the application of EVT in the field of road safety. This approach can predict the probability of a crash consistently and reliably without being dependent on crash data. The EVT approach is applied to estimate the collision risk for uni-variate and bivariate models with the generalized extreme value in the BM approach and the generalized Pareto distribution in the POT approach. The study highlights the different traffic conflicts defined by different types of indicators. However, PET and TTC have shown promising results in previous literature, both PET and TTC are good indicators for near-crash rear-end, lane change, and crossing collisions. In summary, early work argued that the bivariate extreme value model is more useful for estimating safety than the uni-variate model. EVT (POT) approach’s overall performance is more reliable and practical than EVT (BM) approach. However, the researchers argued that more comparative research is required before a firm conclusion can be drawn regarding which technique is superior. The main challenge of Extreme value theory is varying degrees of validity. These groups were determined by analyzing crashes based on historical data and 95% Poisson confidence intervals. However, previous literature has shown that the accuracy of both categories is quite difficult. Underlying conflict process is independent and uniformly distributed, and time stationary. Existing research is typically constrained in terms of data collection regarding traffic conflict observations, study sites, and time intervals that may not adequately cover a variety of traffic. Furthermore, extended observation periods and more sites are also required to address this

650

D. Singh et al.

issue. The multi-variate EVT method may be of interest to improve EVT performance by combining multiple indicators. Aside from that, the implications of the EVT approach in safety analysis are primarily done in homogeneous traffic conditions, where traffic follows a lane-based pattern. Further research is needed to take into account non-lane-based traffic patterns in heterogeneous environments. Future research requires the application of EVT in naturalistic driving datasets to prove more relationships between driver characteristics and accident causes and provides a comprehensive crash analysis. More research must include the influence of geometric and environmental factors on threshold heterogeneity.

References 1. Ahmed I, Puan OC, Ismail CR (2013) A comparative review of road safety audit guidelines of selected countries. Jurnal Teknologi (Sci Eng) 65(3):67–74. https://doi.org/10.11113/jt.v65. 2148 2. Alsop J, Langley J (2001) Under-reporting of motor vehicle traffic crash victims in New Zealand. Accid Anal Prev 33:353–359 3. Borsos A, Farah H, Laureshyn A, Hagenzieker M (2020) Are collision and crossing course surrogate safety indicators transferable? A probability based approach using extreme value theory. Accid Anal Prev 143:105517. https://doi.org/10.1016/j.aap.2020.105517 4. Chauhan R, Dhamaniya A, Arkatkar S (2021) Spatiotemporal variation of rear-end conflicts at signalized intersections under disordered traffic conditions. J Transp Eng Part A: Syst 147(11):05021007. https://doi.org/10.1061/jtepbs.0000589 5. Coles S (2001) An introduction to statistical modeling of extreme values. Springer, London, UK 6. Elvik R (2002) The importance of confounding in observational before-and-after studies of road safety measures. Accid Anal Prev 34(5):631–635. https://doi.org/10.1016/S0001-4575(01)000 62-8 7. Farah H, Azevedo CL (2017) Safety analysis of passing maneuvers using extreme value theory. IATSS Res 41(1):12–21. . https://doi.org/10.1016/j.iatssr.2016.07.001 8. Fisher RA, Tippett LHC (1928) Limiting forms of the frequency distribution of the largest or smallest member of a sample. Paper presented at the mathematical proceedings of the Cambridge Philosophical Society 9. Goyani J, Paul AB, Gore N, Arkatkar S, Joshi G (2021) Investigation of crossing conflicts by vehicle type at unsignalized T-intersections under varying roadway and traffic conditions in India. J Transp Eng Part A: Syst 147(2):05020011. https://doi.org/10.1061/jtepbs.0000479 10. Hyde T, Wright CC (1986) Extreme value methods for estimating road traffic capacity. Transp Res Part B 20(2):125–138 11. Lord D, Mannering F (2010) The statistical analysis of crash-frequency data: a review and assessment of methodological alternatives. Transp Res Part A 44:291–305 12. Meuleners LB, Hendrie D, Lee AH, Legge M (2008) Effectiveness of the black spot programs in Western Australia. Accid Anal Prev 40(3):1211–1216. https://doi.org/10.1016/j.aap.2008. 01.011 13. Moher D, Liberati A, Tetzlaff J, Altman DG (2009) Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. BMJ (Online) 339(7716):332–336. https://doi.org/10.1136/bmj.b2535 14. Pickands J (1975) Statistical inference using extreme order statistics. Ann Stat 3:119–131 15. Salifu M, Ackaah W (2012) Under-reporting of road traffic crash data in Ghana. Int J Inj Control Saf Promot 19:331–339

A Review on Surrogate Safety Measures Using Extreme Value Theory

651

16. Saunier N, Sayed T (2008) Probabilistic framework for automated analysis of exposure to road collisions. Transp Res Rec: J Transp Res Board 2083:96–104 17. Songchitruksa P, Tarko AP (2006) The extreme value theory approach to safety estimation. Accid Anal Prev 38(4):811–822. https://doi.org/10.1016/j.aap.2006.02.003 18. Tarko A, Davis G, Saunier N, Sayed T, Washington S (2009) White paper: surrogate measures of safety. In: Committee on safety data evaluation and analysis (ANB20) 19. World Health Organization (2019) Global action plan on physical activity 2018–2030: more active people for a healthier world. World Health Organization 20. Zheng L, Ismail K, Meng X (2014) Freeway safety estimation using extreme value theory approaches: a comparative study. Accid Anal Prev 62:32–41. https://doi.org/10.1016/j.aap. 2013.09.006 21. Zheng L, Ismail K, Meng X (2014) Shifted Gamma-generalized Pareto distribution model to map the safety continuum and estimate crashes. Saf Sci 64:155–162. https://doi.org/10.1016/ j.ssci.2013.12.003 22. Zheng L, Ismail K, Sayed T, Fatema T (2018) Bivariate extreme value modeling for road safety estimation. Accid Anal Prev 120(January):83–91. https://doi.org/10.1016/j.aap.2018.08.004 23. Zheng L, Sayed T (2019) From univariate to bivariate extreme value models: approaches to integrate traffic conflict indicators for crash estimation. Transp Res Part C: Emerg Technol 103:211–225. https://doi.org/10.1016/j.trc.2019.04.015

Examining the Evasive Behaviour of Pedestrians to Measure Their Degree of Vulnerabilities at Unsignalised Intersections George Kennedy Lyngdoh , Aakash Bhardwaj, Manish Dutta, and Suprava Jena

Abstract Pedestrian safety has become a significant problem in today’s world, and pedestrian casualties have escalated in developing countries where there are no special provisions for the movement of such vulnerable road users. The interaction between pedestrians and vehicles must be prioritised and requires extended study. Therefore, surrogate safety measures (SSMs) come in handy to find the relative spatial and temporal measures of road users under conflict. This study’s primary objective is to identify evasive action-based pedestrian safety indicators that are best suited for predicting pedestrian behaviour under mixed traffic conditions. The pedestrian’s step frequency and lateral deviation are proven useful in measuring evasive actions based on the pedestrian’s trajectory data. Furthermore, an expert-based analysis is undertaken to evaluate which evasive action-based parameters are most appropriate in identifying the severity of pedestrian safety. Hence, it was found that the lateral deviation has a more significant potential influence on severity identification than step frequency. Keywords Pedestrian safety · Surrogate safety measures · Trajectory data · Un-signalised intersection

G. K. Lyngdoh (B) · A. Bhardwaj · S. Jena National Institute of Technology Silchar, Silchar 788010, India e-mail: [email protected] S. Jena e-mail: [email protected] M. Dutta Institute of Technology, Nirma University, Ahmedabad, Gujarat 382481, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_52

653

654

G. K. Lyngdoh et al.

1 Introduction Pedestrians, also known as “vulnerable road users,” are particularly prone to severe injuries and fatalities when they encounter vehicular collisions. Pedestrian safety at junctions is a significant problem that must be addressed effectively. Cyclists and pedestrians are the most vulnerable road users, accounting for 26% of all road accident-related deaths worldwide(WHO) [1]. Pedestrians accounted for 13.8% of road accident fatalities in India in 2017, adding to 20,457 deaths (MORTH) [2]. The heterogeneity in the traffic conditions where multiple vehicles of different dynamic and static characteristics share the same road with pedestrians contributes to the high crash rate. As seen on Indian roads, the loose lane discipline and the poor yielding behaviour of both vehicles and pedestrians further worsen the situation. Traditional road safety methodologies in India are based on accidents reported in the police’s First Information Reports (FIR). These records are very basic, and pedestrian safety analysis using such data of varying quantity and quality will not produce accurate results. Therefore, surrogate safety measures (SSMs) are necessary for safety analysis for a developing nation like India to overcome such problems.

2 Literature Review The pedestrian-vehicle interaction is very dynamic in nature since it depends on numerous factors and conditions that occur when the movement of a pedestrian and a vehicle are in conflict [3]. A traffic conflict is “… an event involving two or more road users, in which the action of one user causes the other user to make an evasive manoeuvre to avoid a collision” [4]. The nature of this conflict may differ based on the geometry of the road where it takes place, the type of intersections, any legal restrictions, and the requirements of the driver and pedestrian [5]. According to Fu et al. [6], the willingness of the driver to yield and the aggressiveness of the pedestrian are just two of the numerous variables that might affect this conflict. It may result in instances where the driver makes an unexpected judgement, while the pedestrian has already considered the driver’s opinion, leading to a conflict or even an accident. An important factor to consider when assessing pedestrian behaviour is the average walking speed. For instance, Goh et al. [7] compared the crossing speed of pedestrians at both signalised and unsignalised crosswalks, and it was found that the speed is higher in signalised crosswalks. Knoblauch et al. [8] found that on wider roadways, pedestrians frequently have higher crossing speed. The evasive behaviours of pedestrians in conflict situations have been investigated by numerous researchers. Malkhamah et al. [9] reported that evasive action is taken by both drivers and pedestrians to avoid conflict at a Pelican crossing. Tageldin et al. [10] compared two surrogate safety indicators to examine evasive action-based indicators and temporal proximity indicators in assessing the seriousness of pedestrian conflicts in heterogeneous traffic situations. Pedestrian evasive behaviours, such as rushing, running, or

Examining the Evasive Behaviour of Pedestrians to Measure Their …

655

stopping, were used by Medina et al. [11] to categorise pedestrian conflict severity levels. Traditionally, field observations and manual measurements have been used to gather pedestrian data, which are time-consuming, labour-intensive, and costly. Computer vision algorithms were successfully used in numerous pedestrian data collection to automatically detect and track pedestrian and vehicular movement in videos and to conduct a thorough safety analysis of pedestrian behaviour [12, 13].

3 Methodology Historical data of accidents and SSMs have been extensively used to characterised vehicle–pedestrian interaction. The present study aims to identify evasive actionbased parameters to find the dependency of interaction behaviour of pedestrians. Therefore, it is critical to select the best indicator to indicate evasive actions of road users. Figure 1 shows the flowchart of safety analysis adopted in this study.

3.1 Data Acquisition Videographic survey was conducted to collect data from a T-intersection in Silchar, Assam, India. The camera was installed at a height from a nearby building to obtain a clear view of the intersection. Recording was done for two hours every day for five days during peak hour traffic. Table 1 contains information on the junction, including vehicle and pedestrian volumes.

Fig. 1 Proposed methodology of the study

656

G. K. Lyngdoh et al.

Table 1 Data for the observed intersection Intersection

Total video duration (hr)

Pedestrian volume (ped/ hr)

Vehicular volume (veh/ hr)

T-intersection

10

410

1800

Fig. 2 Plane calibration using the perspective grid

3.2 Data Extraction The camera data is analysed using an open-sourced, semi-automated tracking software, Kinovea, to extract trajectory data. The following steps are involved in trajectory-based data acquisition: • Coordinate transformation: A perspective grid of a specified length is created on the video using a homography matrix. Figure 2 depicts a snapshot of the video calibration utilising the perspective grid. • Obtaining trajectory data: The track path option was used to select the appropriate vehicle and pedestrian to track both road users. Any noise in the trajectory data is filtered using the Butterworth filter. Care is taken while setting the feature tracker to reduce errors as much as possible.

4 Results and Discussions 4.1 Analysis of the Selected Evasive Action-Based Indicators Step Frequency. It is the number of times a foot touches the surface per unit time [14]. Seismosignal software is used to calculate the step frequency profile based on the speed profile of the pedestrian obtained from the trajectory data, as shown in Fig. 3. The speed profile is used as input for the software.

Examining the Evasive Behaviour of Pedestrians to Measure Their … Speed, m/s

Fig. 3 Speed profile of pedestrian

657

3 2 1 0 0

0.2

0.4

Time, s

0.6

0.8

1

After analysing all the extracted speed profiles of pedestrians, it has been observed that the maximum number of power amplitude peaks lies in the frequency range of 0.5–2.5 Hz. Figure 4 shows the power spectrum profile of the trajectory data. Lateral Deviation. Pedestrians change their paths, along with step frequency, while crossing the street to avoid collisions. For pedestrians, two sorts of route adjustments were observed: one to avoid collision and the other to shorten the trip distance. The first type of route variation was seen before the collision zone, while the second type was observed after the collision zone [15]. The change in path is characterised by the lateral deviation obtained during trajectory data analysis. Figure 5 shows the lateral deviation (LD) calculation assuming that the pedestrian’s ideal path is perpendicular to the origin when the pedestrian decides to cross. The 85th percentile of the maximum lateral deviation was calculated as shown in Fig. 6 and is equal to 1.18 m, which indicates evasive behaviour. 0.25 Power amplitude

Fig. 4 Power spectrum profile

0.2 0.15 0.1 0.05 0 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9 Frequency (Hz)

Fig. 5 Lateral deviation in the pedestrian path

G. K. Lyngdoh et al.

Fig. 6. 85th percentile maximum lateral deviation graph

cum. % of pedestrians

658 120.00 88.24

100.00

100.00

98.32

65.55

80.00 60.00 31.93

40.00 20.00 0.00 0

0.5 1 1.5 Deviation of pedestrians, m

2

2.5

4.2 Analysis Based on Expert Evaluation To see if the estimated evasive indicators were effective in indicating conflict severity, the ranking of two traffic safety experts was used to compare the evasive indicators. The experts involved in traffic safety studies were asked to classify these conflicts based on severity. Using Sayed and Zein’s severity criterion, the experts were asked to rate pedestrian evasive behaviours within each category [16]. Consistency Test. Cohen’s kappa coefficient was used to assess the level of agreement between the two experts [17]. The agreement between the two experts was determined using the classification of both experts, as shown in Table 2 for the pedestrian step frequency dataset and Table 3 for the pedestrian lateral deviation dataset. In Eq. 1, the kappa coefficient is defined as follows: Table 2 The outcomes of two safety experts’ agreement for the pedestrian step frequency dataset Expert 1

Expert 2 low

Medium

High

Low

4

3

0

Total 7

Medium

3

3

3

9

High

0

2

7

9

Total

7

8

10

25

Table 3 The outcomes of two safety experts’ agreement for the pedestrian lateral deviation dataset Expert 1

Expert 2 low

Medium

High

Total

Low

4

4

1

9

Medium

3

3

1

7

High

2

1

6

9

Total

9

8

8

25

Examining the Evasive Behaviour of Pedestrians to Measure Their …

k=

P − Pe 1 − Pe

659

(1)

where P = % agreement observed, Pe = % agreement expected by chance alone. The value of kappa, i.e. k, was determined to be 0.34 and 0.28 for pedestrian step frequency and lateral deviation, respectively. Equation 2 is used to compute Kappa variance, shown as follows. Σ var (k) =

1 × N

)2 (Σ 2 2 P j − P j j j ( )2 Σ 1 − j P j2

(2)

where N = number of cases in total, j = 1 to n = 3 categories of classification, Pj = proportion of all jth category assignments. Table 4 summarises the findings of the kappa test for evasive action indication. Under the hypothesis of no agreement beyond chance and using the central limit √ theorem, the k/ var(k) may √ be roughly distributed as a standard normal variation [17]. It can be seen that k/ var(k) is greater than the Z-value at a 95% significance level (Z = 1.96). Hence, there is a statistically strong agreement between the two experts. Correlation of Ranking. The experts’ ratings and the ranking based on evasive indicators are compared using the Spearman rank correlation. The ranking correlation evaluates the intensity of the ranking of various conflicts with respect to the researched indicators. Equation 3 is used to derive this correlation, and the results are shown in Table 5. Σ 6 d2 (3) r =1− 3 n −n where Table 4 Results of consistency test between experts Evasive indicator pedestrian step frequency

Pedestrian lateral deviation

P

0.56

0.52

Pe

0.34

0.33

k

0.34

0.28

Var(k) √ k/ var(k)

0.0204

0.0203

2.35

1.97

660

G. K. Lyngdoh et al.

Table 5 Spearman rank correlation coefficient based on experts’ rankings Spearman’s rank correlation coefficient

Step frequency

Lateral deviation

Expert 1

0.7088

0.7177

Expert 2

0.6680

0.7861

d = difference between the ranks of observations, n = number of conflicts. Severity Trend. The efficacy of the proposed indicators is examined in this section. The first stage in comparing severity measurements is to examine the trend along the expert-assigned categories. In general, the expert categories’ means exhibited a decreasing tendency from the most severe, i.e. high, to the least severe, i.e. low. The profile in Fig. 7 shows a declining pattern, indicating that the severity ranked by experts is highly correlated with step frequency and lateral deviation.

5 Conclusions This study aims to create a framework for studying pedestrian interactions with vehicles at unsignalised junctions. The significant findings are discussed below: • This study examines pedestrian step frequency and lateral deviation as the evasive action-based indicators based on the literature gap. The selected indicators are effective and dependable in recognising conflicts at an unsignalised junction under mixed traffic situations. • It was discovered that the greatest number of power amplitude peaks are in the range of 0.5–2.5 Hz, indicating the pedestrian step frequency during evasive action. To validate the frequency range obtained from the power spectrum profile, 25 pedestrian profiles were chosen randomly, and their step frequencies were manually calculated. The manually calculated step frequency range is 0.64– 2.27 Hz. Both values are nearly identical, indicating better accuracy of the result. • The estimated evasive indicators were compared to the ranking of two experts to see if they effectively indicated conflict severity. Only 25 pedestrian trajectories were randomly employed because a larger number of pedestrian trajectories from the research area might complicate the expert rating procedure. The lateral deviation is shown to be more suitable for severity identification than step frequency.

Examining the Evasive Behaviour of Pedestrians to Measure Their … Mean of Step Frequency

(a)

1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

Mean of S.F by expert 1 Mean of S.F by expert 2

High 1.50825 1.42578

Medium 0.75955 0.71847

661

Low 0.59989 0.66528

Expert Categories

(b)

Mean of Lateral Deviation

Mean of S.F by expert 1

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Mean of L.D by expert 1 Mean of L.D by expert 2

High 0.6313362 0.6717197

Mean of S.F by expert 2

Medium 0.5274335 0.4963366

Low 0.3646793 0.3679693

Expert Categories Mean of L.D by expert 1

Mean of L.D by expert 2

Fig. 7 a Severity trends of expert categories for step frequency. b Severity trends of expert categories for lateral deviation

References 1. Organization WH (2018) Global status report on road safety 2018: summary. World Health Organization 2. MoRTH (2017) Road accidents in India. Transportation research wing. Ministry of road transport and highways, Government of India 3. Lu L, Ren G, Wang W, Chan CY, Wang J (2016) A cellular automaton simulation model for pedestrian and vehicle interaction behaviors at unsignalized mid-block crosswalks. Accid Anal Prev 95:425–437 4. Parker Jr, MR, Zegeer CV (1989) Traffic conflict techniques for safety and operations: Observers manual. No. FHWA-IP-88-027, NCP 3A9C0093. United States. Federal Highway Administration 5. Schroeder B, Rouphail N, Salamati K, Hunter E, Phillips B, Elefteriadou L, Mamidipalli S (2014) Empirically-based performance assessment and simulation of pedestrian behavior at

662

6. 7. 8. 9. 10. 11.

12.

13. 14. 15. 16. 17.

G. K. Lyngdoh et al. unsignalized crossings. Southeastern Transportation Research, Innovation, Development and Education Center Fu T, Miranda-Moreno L, Saunier N (2018) A novel framework to evaluate pedestrian safety at non-signalized locations. Accid Anal Prev 111:23–33 Goh BH, Subramaniam K, Wai YT, Mohamed AA, Ali A (2012) Pedestrian crossing speed: the case of Malaysia. Int J Traffic Transp Eng 2(4):323–332 Knoblauch RL, Pietrucha MT, Nitzburg M (1996) Field studies of pedestrian walking speed and start-up time. Transp Res Rec 1538(1):27–38 Malkhamah S, Tight M, Montgomery F (2005) The development of an automatic method of safety monitoring at Pelican crossings. Accid Anal Prev 37(5):938–946 Tageldin A, Sayed T, Shaaban K (2017) Comparison of time-proximity and evasive action conflict measures: case studies from five cities. Transp Res Rec 2661(1):19–29 Medina J, Benekohal R, Wang M-H (2008) In-street pedestrian crossing signs and effects on pedestrian–vehicle conflicts at University Campus Crosswalks. Transportation Research Board Annual Meeting, Washington, DC Hediyeh H, Sayed T, Zaki MH, Ismail K (2014) Automated analysis of pedestrian crossing speed behavior at scramble-phase signalized intersections using computer vision techniques. Int J Sustain Transp 8(5):382–397 Zaki MH, Sayed T (2014) Automated analysis of pedestrians’ nonconforming behavior and data collection at an urban crossing. Transp Res Rec 2443(1):123–133 Hediyeh H, Sayed T, Zaki MH, Mori G (2014) Pedestrian gait analysis using automated computer vision techniques. Transp A: Transp Sci 10(3):214–232 Kathuria A, Vedagiri P (2020) Evaluating pedestrian vehicle interaction dynamics at unsignalized intersections: a proactive approach for safety analysis. Accid Anal Prev 134:105316 Sayed T, Zein S (1999) Traffic conflict standards for intersections. Transp Plan Technol 22(4):309–323 Fleiss JL (1971) Measuring nominal scale agreement among many raters. Psychol Bull 76(5):378

Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network Partitions Facilitating Feedback Linearization-Based Control Saumya Gupta, Pushkin Kachroo, Shaurya Agarwal, and Kaan Ozbay

Abstract This paper explores a novel approach to dividing a traffic region (network) into sub-regions for efficient traffic control among the areas. The macroscopic flow diagram (MFD) in each of these sub-regions, referred to as sub-MFD, can then be used to determine the macro-state of that sub-region and subsequently design controllers. The region division is based on the theory of complex networks. We exploit the inherent network characteristics through the PageRank centrality algorithm to identify the most significant nodes in the traffic network. We use these significant nodes as the seeds for a Voronoi diagram-based partitioning mechanism of the network. A feedback linearization-based controller is then presented, which controls the traffic flow between the sub-regions. A case study is performed for the Manhattan area in New York City to demonstrate the network partitioning approach; the control approach is demonstrated through a toy example containing two sub-regions. Keywords Macroscopic Flow Diagram (MFD) · Sub-MFD · Traffic flow control

1 Introduction Efficient traffic control of large-scale urban transportation networks remains a big challenge for researchers. Challenges include uncertainty in user behavior, accurate estimation of origin–destination (OD) trip matrices, and a reasonable estimate of a network state. Traditionally, the developed control algorithms focused on individually controlling each link and signalized intersections. There are multiple probS. Gupta (B) · S. Agarwal Department of Civil, Environmental, and Construction Engineering, University of Central Florida (UCF), Orlando, FL, USA e-mail: [email protected] P. Kachroo Department of Electrical and Computer Engineering, University of Nevada Las Vegas (UNLV), Las Vegas, NV, USA K. Ozbay Civil and Urban Engineering Department, New York University (NYU), New York, NY, USA © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_53

663

664

S. Gupta et al.

lems associated with the existing network control strategies. Firstly, it is complex to model traffic dynamics at each link and intersection on a large-scale urban network. Secondly, these models or simulation scenarios are challenging to calibrate due to the high amount of stochasticity involved in traveler behavior and network parameters. Thirdly, even if we develop a realistic model or simulation scenario, many input/output variables associated with the control design make the implementation infeasible. Issues such as controllability and observability of the network come into play as well [1–3]. In addition to link-level control, a macro approach to managing traffic flow between regions is also needed in large-scale networks. Researchers have been exploring the idea of performing perimeter control of an area based on an aggregation of traffic conditions in a network. This approach helps in reducing the complexity of the modeling and control design, as well as in the deployment of traffic control strategies. Aggregating traffic states over a region has recently been of active interest to researchers due to the desire to efficiently manage traffic in larger regions. The concept of the Macroscopic Fundamental Diagram (MFD) has been around for a while but has only recently started to gain significant traction and attention from researchers. Godfrey presented the idea of MFD through his research in 1969, where he reported the network-wide relationship between observed speed and density [4]. Herman and Prigogine [5] presented the idea of aggregation of traffic variables for a city, where they consider moving vehicles and the fraction of vehicles that are stopped to estimate aggregated traffic conditions in a traffic network. Proof of the existence of the MFD using analysis of experimental data from downtown Yokohama, Japan, was presented in [6]. The authors further made a case for the MFD relation between average speed and densities over the network of a certain class and provided some analytical treatment as well in [7]. A generalized MFD that uses the inhomogeneity of the traffic has been proposed in [8]. As an example of the importance of MFD and its usage in practical problems, [9] used it for the model predictive perimeter control of an area partitioned as two aggregated regions. Geroliminis et al. also developed an optimal perimeter control problem for two-region urban cities by utilizing the MFD concept in [10] and solved it by using a model predictive control approach. Another study based on using a model predictive control is [11]. Daganzo developed an adaptive control approach to improve urban mobility and decrease congestion in [12] by observing and controlling aggregate vehicular accumulations in neighborhoods. Ekbatani et al. also exploited the idea of a network fundamental diagram (NFD) to improve mobility in congested conditions by applying gating techniques using feedback control [13]. Although the concept of MFD is promising and has a lot of potential, its existence is not guaranteed. A well-defined MFD exists only under certain conditions, which were discussed in [14]. However, due to the large network size and variations within the network, aggregated variables sometimes do not truly represent the entire network. Recent research indicates that link density heterogeneity is very crucial in determining the shape and scatter of MFD [15]. Moreover, heterogeneity can also cause hysteresis loops and degradation of network performance [16]. An MFD is expected to be well-defined

Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network …

665

under the condition that the network is homogeneous with similar link properties. However, in reality, large-scale urban networks are expected to have various congestion levels. Now, in order to exploit the usefulness of MFD in designing macroscopic control strategies, it becomes essential to subdivide the network into smaller regions for the existence of sub-MFDs in each of these subregions. The first attempt in this direction was made by Ji & Geroliminis in [17], focusing on clustering of the network based on spatial congestion distribution for a given time period. Contribution: This paper aims to explore the inherent network characteristics to create effective sub-division algorithms. We present a novel methodology to divide an urban network into sub-networks, where an MFD determines the macro state of those sub-networks. The region division is based on the theory of complex networks [18]. We exploit the inherent network characteristics through the PageRank algorithm [19] to identify the most significant nodes in the traffic network. We then use these significant nodes as the seeds for a Voronoi diagram-based partitioning mechanism of the network [20, 21]. A feedback linearization-based controller is then presented, which controls the flow among these sub-regions. A case study is performed for the Manhattan area in New York City, and results are provided through simulations.

2 Background 2.1 Complex Networks Cyber-physical systems are becoming the backbone of modern society. These systems have a very high level of connectivity and the study area of complex networks has been developing to answer many questions in this field. Many introductory survey papers [18, 22] and several books exist that provide details of the theory and applications of complex networks [23–25]. Different types of networks that come under this study area include computer networks, social networks, economic networks, biological networks, and transportation networks. This paper uses the theory of complex networks to perform dynamic aggregation of regions so that MFD-based control strategies can be applied to the overall network. Specifically, we use the page rank algorithm to identify the most important nodes in a transportation network. This can be performed based on the structure of a directed graph. It can also be performed on a weighted directed graph where the weights represent the dynamic traffic conditions, for instance, in terms of time-varying traffic densities, on the links of the graph.

666

S. Gupta et al.

2.2 PageRank Algorithm Google uses the PageRank algorithm to rank web pages in their search results [19]. This algorithm measures the importance of web pages by counting the number and quality of links. Next, we will provide some basic terminologies and discussion on the subject. Consider a non-weighted network (.N , .E), where .N is the set of nodes and .E is the set of edges, such that N = {a, b, c, d, . . .} , E = {(a, b) |

.

∀ pair of connected nodes} Now the basic node centrality/significance index can be defined in terms of the degree of the node. For instance, the degree of node .a is representative of the direct number of links it has and is defined as degree (a) = |{(a, b) ∈ E}|,

.

(1)

where .|.| is the degree operator giving the total count of valid connections of the node .a. Another idea immediately following the degree centrality concept is that node centrality is proportional to the cumulative degree of its direct neighbors. Mathematically speaking: Σ .nodecentrality (a) = degree (b), (2) (a,b)∈E

The underlying argument in the PageRank algorithm is that the weight distribution of each connection is not uniform. Rather, it is inversely proportional to the number of other connections that the neighboring nodes have. The following recursive formula defines PageRank centrality: PageRank (a) = α

.

Σ PageRank (b) 1 − α + degree (b) n (a,b)∈E

(3)

where .0 < α < 1 is the damping parameter, and .n is the total number of nodes.

2.3 Voronoi Diagrams Voronoi diagrams provide a computational geometric method to divide a region into subregions based on distance from a given set of seed points [20, 26]. The problem with the Voronoi diagram is as follows:

Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network …

667

[Voronoi Problem] Given .Ω ⊂ Rn , and .si ∈ Ω, .i ∈ {1, 2, ·, N }, then find .vi , .vi ⊂ Ω, .i ∈ {1, 2, ·, N }, such that N | | .

vi = Ω, vi0 ∩ v 0j = ∅

and

i=1

vi = {x ∈ Ω | d(x, si ) ≤ d(x, s j ), ∀ j /= i} Here, the subscript on the regions such as .vi0 means the interior of the set, and the function .d(x, y) is the distance between the points .x and . y in the metric space.

3 Main Algorithm The flowchart for the algorithm for the overall subdivision, aggregation, and region feedback control is shown in Fig. 1. The task of creating subregions based on the PageRank algorithm can be performed once in a static setting or can be repeatedly applied as the traffic in the system changes, which can produce changing boundaries of various regions in a time-varying fashion. The static map data (link lengths, etc.) can be used as an input in the static setting to perform the PageRank algorithm and obtain the most critical nodes of the directed graph (digraph). Alternately, in a dynamic environment, one can create a weighted digraph, where the weights on the links are obtained via traffic densities. This would give a changing set of the most critical nodes as the traffic conditions change. After obtaining the nodes, we apply the Voronoi algorithm to divide the overall region into subregions. Once the regions are obtained, we can use MFD for each region in order to develop a control law on the simplified network model.

4 Case Study Once the subregions have been established with their corresponding MFDs, we write down the conservation flow dynamics for traffic densities for each subregion being its state variable. This provides us with state space control dynamics where control variables are also used to indicate how traffic control has to be performed. For instance, if

Data (map, traffic)

PageRank Algorithm

Fig. 1 Main algorithm flowchart

Significant Nodes

Voronoi Diagrams

sub regions

Sub-region Control using MFD

668

S. Gupta et al.

(a) Top Nodes

(b) Top Nodes

(c) Subnetworks

Fig. 2 Manhattan, New York area divisions

the perimeter control in the region has to be performed using a ramp control [27] or a gating mechanism, a corresponding control variable will be present in the dynamics. Once the dynamics are established, one can design a feedback control law to satisfy some design requirements, such as some asymptotic performance or optimality. An example of dynamics using MFD is given in [10]. We consider the Manhattan road network for the application of the proposed approach and divide it into sub-networks based on the PageRank algorithm. The Manhattan Open Street Map network was downloaded from www.openstreetmap.org. Which was then further parsed and processed to extract the valid nodes (intersections) and their latitude-longitude values. The total number of valid nodes in the data for the Manhattan road network was .446, 415. We next analyze the data using NetworkX package in python and apply the PageRank algorithm to obtain the ten most significant nodes. Figure 2a shows the top significant nodes marked on the Manhattan road network. Sets of nodes that were very close and overlapping were considered as one. Figure 2b shows the markings of nodes on a Google map. It is pretty interesting to note that most of these nodes are close to the major entry bridges/tunnels into the Manhattan area. We next apply the Voronoi diagram method to divide the network into sub-networks, using these significant nodes as seeds. The divided regions are shown in Fig. 2c. The subdivided network can now be represented as a weighted digraph to generate a state space representation of finite dimensional traffic dynamics for the entire network. The directed graph is shown in Fig. 3. In fact, the output of the Voronoi diagram can lead to more features, such as extra subregions, depending on the local geography and traffic density distribution. For instance, in this case, the waterway further divides the nodes.2 and nodes.5. Each MFD region should have a low variance in its internal traffic density distribution. Hence, the region boundary design can be performed by combining the results of the Voronoi algorithm, geographical features,

Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network … Fig. 3 Network digraph

n3 n1

669

n4

n2

n6

n7

n5

and traffic density variance. We can modify the digraph based on these additional changes. We now develop the dynamics for the system for the traffic moving on the digraph in Fig. 3. Control variables can appear at multiple places in the equations depending on what actual mechanism is available in the system. For instance, congestion pricing might be used to control the inflow into a toll road/bridge; other control mechanisms may include ramp metering or adaptive signalized intersections. Control objectives can be created based on the desired values of the MFD variables in the system dynamics.

5 Control Design for Two Region Case An example of dynamics using MFD is given in [10]. Borrowing the dynamics from there, we have its labeled digraph and equations in (4) (Fig. 4). dn 11 (t) dt dn 12 (t) dt . dn 21 (t) dt dn 22 (t) dt

= q11 (t) + u 21 (t)M21 (t) − M11 (t) = q12 (t) − u 12 (t)M12 (t) (4) = q21 (t) − u 21 (t)M21 (t) = q22 (t) + u 12 (t)M12 (t) − M22 (t)

The region .r1 has .n 1 total number of vehicles which is a sum of .n 11 , the number of vehicles in .r1 with destination in .r1 , and .n 12 , the number of vehicles in .r1 with destination in .r2 . The variables for region .r2 are defined analogously. The variable .qi j indicates the trips starting from region .i with destination in region . j. The variable

Fig. 4 Two region network

M11 q11

M12 u12

r2

r1 q12

q21

M21 u21

M22

q22

670

S. Gupta et al.

Mi j indicates the trips ending in region.i with origin in region. j. The control variables u ∈ [0, 1] indicate the fraction of the flow . Mi j they allow to go through. The flow variables . Mi j are related to .n i , .n i j , and .G i (n i ) where .G i (n i ) is the MFD variable indicating the average flow in a region as a function of the total number of vehicles (the product of the average density in a region and the total lane length in the region) in that region. These relationships are noted in Eq. (5).

.

. ij

.

n i (t) = n ii (t) + n i j (t), i /= j ,

and

Mi j (t) =

n i j (t) G i (n i ) n i (t)

(5)

In order to design control laws to try to achieve desired flows obtained by maintaining critical densities (or number of vehicles) in each region, we combine Eq. (4) with Eq. (5) to obtain ] [ ] [ ][ ] [ d n 1 (t) q11 (t) + q12 (t) − M11 (t) M21 (t) −M12 (t) u 21 (t) . (6) = + q21 (t) + q22 (t) − M22 (t) −M21 (t) M12 (t) u 12 (t) dt n 2 (t) We rewrite Eq. (6) using matrix notation which is implied by that equation as .

dn(t) = Q(t) + M(t)u(t) dt

(7)

Equation (6) is decoupled by the control law: u(t) = M −1 [−Q + v]

.

(8)

where the vector .v has components .v1 and .v2 that are chosen as v (t) = −K i (n i (t) − n i d )

. i

(9)

In order to make the vector .n(t) with components .n i (t) follow a desired vector of number of vehicles in each region given by the vector .n d which has components .n i d , we choose positive values for .ki the control gains. This control is designed to ensure .n(t) → n d (t) as .t → ∞. Figure 5 shows the simulation results for two regions. Values of . K 1 and . K 2 are fixed at .1 and .2, respectively. The simulation results show that both the sub-regions achieve the desired state (number of vehicles) by using the proposed control law. Error term also goes to zero as .t → ∞.

5.1 Hierarchical Control The MFD framework automatically lends itself to a natural hierarchical control structure. Each MFD region should have a low variance traffic density distribution for the MFD to be valid in that region. So, at the higher level, MFD can be used to control

Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network … n1

4

# of Vehicles

8

x 10

1

7.8

−2000

0.8

7.6

−4000

0.6

7.2 7

Region 1

0

1

6

2 3 Time (Hr)

4

5

−6000

0.4

−8000

0.2

−10000

0

1

4

5

0

0

1

−500

5.9

−1000

4

5

4

5

12

0.4

0

5.95

3 2 Time (Hr) Control action u

2

2

x 10

2 3 Time (Hr) Error E

n

4

# of Vehicles

Control action u21

Error (e1) 0

7.4

671

0.3 0.2 0.1

5.85

5.8

−1500

Region 2

0

1

3 2 Time (Hr)

4

5

−2000

0

0

1

3 2 Time (Hr)

4

5

−0.1

0

1

3 2 Time (Hr)

Fig. 5 Simulation results

an inflow, for instance, at a bridge to the entire region. Similarly, at a lower level, the traffic signals and ramps inside the region can be used to maintain a smooth flow, creating a uniform traffic density in that region.

6 Conclusions This paper explored a new approach for creating subregions for an area for traffic control based on the application of the PageRank algorithm from the complex networks theory to identify important nodes, followed by the application of the Voronoi diagram algorithm. The approach then used MFD for each subregion and used feedback control design on the simplified network model. An example problem was studied illustrating these steps applied to that problem, followed by a simulation performed in a two-region network that uses a novel feedback linearization control for perimeter control. The results presented in the paper are preliminary and require a large-scale case study to validate the outcomes. Furthermore, we did not perform a check to validate the existence of MFDs and sub-MFDs in the regions, which could be a potential future research direction.

References 1. Agarwal S, Kachroo P (2019) Controllability and observability analysis for intelligent transportation systems. Transp Dev Econ 5(1):1–10 2. Contreras S, Kachroo P, Agarwal S, Observability and sensor placement problem on highway segments: A traffic dynamics-based approach. IEEE Trans Intell Transp Syst

672

S. Gupta et al.

3. Agarwal S, Kachroo P, Contreras S, A dynamic network modeling-based approach for traffic observability problem. IEEE Trans Intell Transp Syst 4. Godfrey J (1969) The mechanism of a road network. Traffic Eng Control 11(7):323–327 5. Herman R, Prigogine I (1979) A two-fluid approach to town traffic. Science 204(4389):148–151 6. Geroliminis N, Daganzo CF (2008) Existence of urban-scale macroscopic fundamental diagrams: some experimental findings. Transp Res Part B: Methodol 42(9):759–770 7. Daganzo CF, Geroliminis N (2008) An analytical approximation for the macroscopic fundamental diagram of urban traffic. Transp Res Part B: Methodol 42(9):771–781 8. Knoop VL, van Lint H, Hoogendoorn SP (2015) Traffic dynamics: its impact on the macroscopic fundamental diagram. Phys A: Stat Mech Its Appl 438:236–250 9. Haddad J, Ramezani M, Geroliminis N (2012) Model predictive perimeter control for urban areas with macroscopic fundamental diagrams. In: American control conference (ACC). IEEE, pp 5757–5762 10. Geroliminis N, Haddad J, Ramezani M (2013) Optimal perimeter control for two urban regions with macroscopic fundamental diagrams: a model predictive approach. IEEE Trans Intell Transp Syst 14(1):348–359 11. Deo P, De Schutter B, Hegyi A (2009) Model predictive control for multi-class traffic flows. In: Control in transportation systems, pp 25–30 12. Daganzo CF (2007) Urban gridlock: macroscopic modeling and mitigation approaches. Transp Res Part B: Methodol 41(1):49–62 13. Keyvan-Ekbatani M, Kouvelas A, Papamichail I, Papageorgiou M (2012) Exploiting the fundamental diagram of urban networks for feedback-based gating. Transp Res Part B: Methodol 46(10):1393–1403 14. Geroliminis N, Sun J (2011) Properties of a well-defined macroscopic fundamental diagram for urban traffic. Transp Res Part B: Methodol 45(3):605–617 15. Daganzo CF, Gayah VV, Gonzales EJ (2011) Macroscopic relations of urban traffic variables: Bifurcations, multivaluedness and instability. Transp Res Part B: Methodol 45(1):278–288 16. Geroliminis N, Sun J (2011) Hysteresis phenomena of a macroscopic fundamental diagram in freeway networks. Transp Res Part A: Policy Pract 45(9):966–979 17. Ji Y, Geroliminis N (2012) On the spatial partitioning of urban transportation networks. Transp Res Part B: Methodol 46(10):1639–1656 18. Newman ME (2003) The structure and function of complex networks. SIAM Rev 45(2):167– 256 19. Page L, Brin S, Motwani R, Winograd T (1999) The pagerank citation ranking: Bringing order to the web. Stanford InfoLab, Technical Report 1999–1966, previous number = SIDL-WP1999-0120 20. Aurenhammer F (1991) Voronoi diagrams-a survey of a fundamental geometric data structure. ACM Comput Surv (CSUR) 23(3):345–405 21. Okabe A, Boots B, Sugihara K, Chiu SN (2009) Spatial tessellations: concepts and applications of Voronoi diagrams, vol 501. Wiley, New York 22. Newman M, Barabasi A-L, Watts DJ (2006) The structure and dynamics of networks. Princeton University Press 23. Cohen R, Havlin S (2010) Complex networks: structure, robustness and function. Cambridge University Press 24. Newman M (2010) Networks: an introduction. Oxford University Press 25. Dorogovtsev SN (2010) Lectures on complex networks, vol 24. Oxford University Press Oxford 26. Agarwal S, Sancheti A, Khaddar R, Kachroo P (2013) Geospatial framework for integration of transportation data using voronoi diagrams. In: Transportation research board 92nd annual meeting, no. 13-5378 27. Agarwal S, Kachroo P, Contreras S, Sastry S (2015) Feedback-coordinated ramp control of consecutive on-ramps using distributed modeling and godunov-based satisfiable allocation. IEEE Trans Intell Transp Syst 16(5):2384–2392

Meta-Analysis of the Methodologies Used for Road Accident Costing and Conceptualizing Framework for Road Accident Compensation Adil Ata Azmi and Sewa Ram

Abstract Road accidents have long-term negative consequences for society and have significant economic impacts, there are several techniques for valuing road accident-related expenses. Those from lower socioeconomic groups are more likely to be involved in accidents. In India, there is a wide range of accident scenarios at the state and city levels. About half of India’s union territories and states (45%) have a fatality risk that is higher than the national average. Majority of developed and emerging countries are focusing on road safety to reduce fatalities. In India, approximately 1.5–1.6 lakh people die each year as a result of vehicle accidents. As a result, it requires immediate attention in order to take the essential steps to improve the deteriorating situation by exploring and implementing the best method which can value the detrimental impacts of road accidents. This study will try to collate the different methods of road accident costing in literature, based on their advantages, disadvantages, and applicability, and will formulate a framework base, on the most suitable method for developing countries, with a focus on India. This paper will also address the key component of traffic accidents and compensation. Keywords Meta-analysis · Methodologies · Accidents · Costing · Compensation

1 Introduction The human population is ever-increasing and so is the need to travel, As the world’s population grows, so does the number of automobiles on the road. As per [1], India has about 300 million vehicles. The number of automobiles has increased at an exponential rate, WHO [2, 3]. As per [4], a road accident is defined as an “Unplanned event in a chain of planned or controlled events”. While Deleon et al. [5] have defined A. A. Azmi (B) · S. Ram Department of Transport Planning, School of Planning and Architetcure, New Delhi, India e-mail: [email protected] S. Ram e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_54

673

674

A. A. Azmi and S. Ram

a road accident as error with sad consequences. Most, if not all road accidents are a consequence of human error and come with a human cost attached to them [5]. As per the stats of World Health Organization (WHO), road accidents claim the lives of an estimated 1.3 million people each year, worldwide. Non-fatal injuries are estimated to affect 20 million to 50 million people worldwide, with the majority of them becoming disabled [2]. According to WHO projections for India, for the year 2019, 1,54,732 deaths and 4,37,396 injuries were reported annually [2]. According to Ministry of Road Transport and Highways (MoRTH), 1,50,785 people were killed and 4,94,624 people were injured in India in the year 2018 [6]. As per [7], India has the highest number of fatalities due to road traffic accidents as compared to any other country, which also confirms the finding by [8, 9]. Another estimate predicts that, in India, the deaths due to road accidents are expected to cross the mark of 250,000 by the year 2025 if no immediate, preventive measures are taken [10]. Road accidents and deaths vary by age, gender, month of the year, time, and even social background in India. The 30–59-year-old age group is the most vulnerable. Males are more vulnerable to traffic accidents than females [10]. Accidents are also more likely to occur during times of extreme weather and during working hours. Although mobility due to transportation systems has made our daily lives easy but it does attach the cost of lives lost as a result of traffic accidents. Study by [5] referred to vehicles as “High-velocity moving lumps of metal” and their impacts as “Weapons of mass destruction” [11]. The deaths due to road accidents cost India about 1.5 lakh lives annually, which is rarely discussed [6]. Even if they are discussed, their scale and impacts are very less, and due to this reason, road accident deaths can also be referred to as a “Hidden Pandemic” as per [12]. The research paper is focused to draw attention to this “Hidden pandemic”. This research will highlight different methods used for road accident costing and will try to formulate the best suitable method for Indian conditions. The next section, of the research paper, discusses the methodology, followed by, the detrimental socio-economic effects of road accidents, and economic models that are currently used with the pros and cons in detail in the subsequent sections. This research paper is an attempt to highlight the gaps in the literature and to come up with a framework to address the research gap in the Indian context. This paper will also address the important aspect of compensation. Figure 1 shows the death rate and number of vehicles over the years.

2 Research Methodology and Data Collection Various research papers were collected relevant to the present research. Research papers were collected by searching journals from open-source web databases with keywords like road accidents, costing mechanism, compensation, meta-analysis, methodology, accidents, etc. It included published journals, reports by national and international organizations, thesis, and books. The keywords were searched from different academic open sources like Scopus, TRID, WOS, PubMed, etc. This paper is based on a thorough review of more than 304 such articles that were initially

Meta-Analysis of the Methodologies Used for Road Accident Costing …

675

Fig. 1 The first graph shows the number of death and death rate across the globe while the second graph is a relation between the number of vehicles and the death rate worldwide [2]

identified, out of those articles 290 were selected and duplication of the article was avoided, out of the 290 articles 166 were not directly related to the topic, and out of 124 relevant articles, 11 journals could not be assessed and the permission for full text was denied, majority of them were case studies from China. A total of 107 articles from the literature were short-listed after reading the full texts. 95 articles have been mentioned in the reference section and 63 of them have been cross-referred in this article, as and when required. The flowchart for shortlisting the works of literature has been shown in Fig. 2. Figure 3 shows the year-wise publication included in the article while Fig. 4 shows the publishers, whose articles are included in the research paper. The bibliometric network of the keywords has been shown in Fig. 5, along with the domains from where articles were included in the research paper like Road Safety, Traffic Engineering, Economics, Medical, Civil Engineering, Accident Analysis & Injury Prevention, Social and Applied Science. Conventional analysis with systematic literature review (SLR) was done using PRISMA analysis, [5, 9, 13, 14] the methodology followed was to read the literature,

Fig. 2 Flowchart showing how the literature were shortlisted

676

A. A. Azmi and S. Ram

Fig. 3 Year-wise publication of the articles

Fig. 4 Publisher of articles included in the research paper

highlight the key points and add them to an Excel spreadsheet, along with the author’s name, Journal/conference name, year of publication, etc. Based on this, papers with similar conclusions were clubbed together and the results were produced. On the basis of this, the gaps in the previous research have been pointed out along with the future scope of work. Figure 6 shows the country-wise contribution of the research paper to the present article.

3 Accident Scenarios 3.1 Accident Scenario Across the Globe and in India There have been a lot of cases where the accident victim’s life was completely devastated after an accident because of his or her dependence on others and restricted mobility. In case the victims succumb to a fatal injury then it is the family members of the deceased who have to cope through phases of trauma and mental torment. Ultimately, it causes huge monetary loss to the nations, and then there are some losses that are intangible that cannot be gauged precisely in monetary terms. About 70% of the victims who lost their lives in road accidents are from developing countries and, out of them, pedestrians account for 65% of the victims, with minors accounting for 35% of those killed on the highways [15]. Based on reports WHO 2018 report [2], road accidents were among the top 10 (at position 8) causes of deaths globally in 2016. Road injury was the number one cause of death due to “Injuries”. Research by

Meta-Analysis of the Methodologies Used for Road Accident Costing …

677

Fig. 5 Analysis of keyword co-occurrence network and domains from where the papers are included

Fig. 6 County wise contribution of the research papers

678

A. A. Azmi and S. Ram

[16] suggests that the number of people killed in automobile accidents each year (last 5 years) is estimated to reach 5 million, while the count of people injured could be as high as 50 million [16]. Traffic crashes are also expected to rise from the ninth to the fifth largest cause of death by year 2030, resulting in about 2.4 million deaths per year unless quick action is taken [16]. Road accidents result in a significant amount of economic loss both on a personal and national level [17] Apart from the safety issue, it creates health, social and economic problems as well, not only for the victim but also for the family members of the victim [18, 19]. The majority of these fatalities occur in developing nations (74%) and undeveloped countries (16%), which is both a public health and socioeconomic development issue [20]. In India, every hour, there are 15 fatalities and 53 injuries due to road accidents [18]. While in many countries, the situation is improving, the situation in India is actually worsening as claimed by Sahu [18]. Tunnels are one of the most vulnerable accident locations, the drivers often become so focused while entering the tunnel that they completely miss out on the information that are provided by the signages and- markings. The rate of accidents is highest at the beginning of the tunnel and it diminishes as one proceeds further in the tunnel [21]. It has been pointed out that apart from tunnels intersections are the most vulnerable points of conflict, and the severity of accidents increases as and when the volume of vehicles and the angle of collision increases [22–26]. Research favours the notion that the identification and treatment of black spots is one of the options for lowering accident costs [27, 28]. Analysing the road accident pattern by mapping the previous accidents can significantly help in future planning and designs of roads that will help in reducing accidents and the costs attached to it [13].

3.2 Types of Road Accidents and Losses Attached to Them A road accident death is defined in India as “death(s) occurring within 30 days of a road traffic accident”. Fatal accidents, serious injury accidents, major injury accidents, grievous injury accidents, minor injury accidents, and no injury accidents (Property damage only accident) are the six types of road accidents [6]. “A fatal accident happens when one or more people are killed as a result of the accident within 30 days of the accident occurring” [6]. “A serious injury accident is defined as one in which a person is detained in a hospital as an “in-patient” or one in which a person sustains any of the following injuries, whether or not he or she is detained in a hospital: fractures, concussions, internal injuries, crushing, severe cuts and lacerations, or severe general shock requiring medical treatment and the victim requires ICU admission”, “A major injury accident is defined as either a person being detained in a hospital as an “inpatient,” or if any of the following injuries are sustained whether or not he or she is detained in the hospital: fractures, concussions, internal injuries, crushing, severe cuts and lacerations, or severe general shock requiring medical treatment but not ICU admission. The grievous accidents include serious and major injury accidents, While the “Minor accident is one in which there are no deaths or serious injuries but a person is slightly injured. This will be an injury of minor nature such

Meta-Analysis of the Methodologies Used for Road Accident Costing …

679

as a cut, sprain, or bruise, where only first aid is required and does not require hospitalization”. A damage-only accident, also known as property damage only (PDO) accident, occurs when no one is hurt but the vehicle and/or property are damaged. Many academics have previously stated that while road accidents cause economic loss on a personal and national level, the effects of road accidents also have physical and emotional ramifications [17, 18, 29, 30]. Psychiatric symptoms and illnesses are common after grievous or minor traffic accidents, according to Richard, Bridget, and Robert. Apart from safety issues, accidents have become health, social and economic problem as well [18, 19]. An effective strategy to save life losses due to road accidents is by providing emergency medical services to the victims, however, too many centres often reduce the quality of service and it rather increases inefficiency in the system [31]. Thus, it is important to provide timely medical service within the “Golden hour”. 15–44 years are the most productive age group [2]. Men have 2.5 times more chances of an accident as compared to women [24, 32]. In all sorts of accidents, the 16–25 age group bears the greatest negative impacts of a road accident [11, 30, 33–37]. The most venerated age group is 30–59 years old [10]. In lowincome nations, the probability of fatality from road accidents is three times higher than in high-income countries [31]. Ironically, countries with a poor safety record and greater death rates place a low monetary value on victims, whereas countries with a better safety record place a higher monetary value on victims [38]. Because of a lack of statistics, the economic loss caused by accidents in poor nations is frequently overlooked [29]. The lower-middle-class section of society reports the maximum number of fatalities, and this statement is true for all countries. The social class of the parents has a strong relationship with their children meeting with an accident [39, 40]. To assess the economic losses due to accidents, five major studies have been conducted in India previously apart from various other research papers on this topic. A brief of each of the previous studies has been given below.

3.3 Major Economic Studies Conducted in India as a Case Study The first major study was conducted in the year 1982, sponsored by the World Bank as the road user cost study. This was the first large road user cost study, and it was based on data from the city of Delhi. Medical bills, legal fees, property damage, insurance costs, and lost output due to death were all included in the costs. The cost of road accidents was projected to be 0.29% of GDP. The Human capital (HC) cost method was used. In the year 1995, Transport Research Laboratory (TRL) did a study under the name of Overseas Road Note 10 for costing road accidents in developing countries. It did surveys in four countries across the world, including India. Suggested a method for calculating accident costs for developing countries. The methodology was loosely based on the HC method with a slight modification, and recommended it as a method to be used in developing countries. In India, the Accident Cost Study was

680

A. A. Azmi and S. Ram

sponsored by the Ministry of surface transport (MOST), which was titled Research Scheme R-79, in the year 2000. It used Gross Output Approach to calculate the value of road accidents. Data coverage included Insurance data from 16 cities across India and one government hospital data. The economic cost of road accidents in India was estimated to be 0.69% of GDP. Around the same time in the year 2000, another study was conducted as a part of the Sundar Committee of India. The study was outsourced to Tata Consultancy Services (TCS), which was the Working group estimate. It is estimated a loss of around 3% of the country’s GDP or about $550,000 million in 2000, or $22,08,021 million in 2022, taking into account previous inflation rates. Till today many evaluations of road accident costs are made based on this report, along with using the wholesale price index. The latest work on road accident costing in India (was done by S. B. Paul, Transportation Research and Injury Prevention Program (TRIPP), Indian Institute of Technology Delhi (IIT-D), and Delhi Integrated MultiModal Transit System (DIMTS) [41]. 20 state capitals were included in the case study (2 trauma centres per city were surveyed). 250 samples from each city hospital were collected, overall, 5000 samples were collected. Calculated the cost of fatal accidents in India as |9.1 million per fatality, which is the social cost of an accident.

4 Review of Methodologies for Road Accident Costing 4.1 Timeline of the Methodologies Used for Road Accident Costing A timeline of the various methods for road accident costing has been given in Fig. 7. It is evident that studies on traffic accidents and their costs did not become sensitive until the 1960s. The most common and most widely used method was the HC method. In the 1970s, it got criticism from various economists as they claimed that its mechanism is inconsistent with the cost–benefit analysis (CBA) principles, in 1982, first road user study was conducted by WHO and it used the HC method. In the 1990s, the developed countries began shifting toward the willingness to pay (WTP) method as they found the method to be more theoretically sound. In 1995, Transport Research laboratory (TRL) conducted another study and suggested a mechanism for the calculation of the cost of road accidents, which was loosely based on the HC method. In 1995, various other methods were explored and in the early 2000s, the researchers suggested that socioeconomic factors should also be considered while calculating the cost. In the 2020s, hybrid methods are being used to assess the cost of road accidents although the number of case studies are limited as it is a fairly new method.

Meta-Analysis of the Methodologies Used for Road Accident Costing …

681

Fig. 7 Flowchart of the various methodologies for the evaluation of road accident cost

4.2 Costing The values of economic losses come out to be different as different studies have taken different parameters under consideration along with the different methods of calculating the road accident cost. Some of the most common methods that are most commonly used across the globe are namely: human capital approach (HC), the life insurance approach, the court award approach, the implicit public sector valuation approach, the values of the risk-change approach or the willingness-to-pay approach (WTP), and the hybrid method. The different methods used across the globe have been shown in Fig. 8. Hills and Jones-Lee [42] suggest that only two road traffic accident costing procedures, the Human Capital (HC) method and the Willingness to Pay (WTP) method, appear to be directly relevant among these methods. When examining a country’s wealth, the HC approach (Gross Output method) is appropriate. If social welfare is a concern, the willingness to pay method is more suited. The Gross Output technique is recommended by TRL (Transport Research Laboratory, UK) to cost road accidents in developing nations like India. Professionals working

Fig. 8 Timeline of road accident costing methodologies

682

A. A. Azmi and S. Ram

in this field, such as E. Hauer, believe that putting a monetary value on human life is unethical [43]. Miller [44] divided the losses resulting from traffic accidents into the following categories as shown in Fig. 9 [45–48]. This method (HC) assumes an arbitrary value for the loss of grief and pain. Economists also argue that this is not an appreciable method to gauge grief loss and pain. It is more suitable when the priority is national welfare and macroeconomic parameters are to be considered. In the 1970s, HC method was criticized for not being consistent with CBA (Cost– Benefit Analysis) principles. But it is one of the easiest and most widely used methods for developing countries. The majority of the studies conducted in the Indian context have used this method. The victim’s future consumption is subtracted from the gross output value in the net output approach. It can also be stated that the cost of an accident is equal to the “gross production” amount less the discounted worth of the victim’s consumption. Because the future consumption of persons killed in road accidents is deducted, the Net output model has a more conservative economic cost to society. The approach was originally used to evaluate the total net increase to a country’s stock of wealth created through production during an accounting interval, rather than to calculate the economic loss due to accidents. This method is used for accident cost calculation in limited studies, it is difficult to find studies in the Indian context that uses net output method. The life insurance technique calculates the total cost of real resources and the price that average people are ready to pay to guarantee their own lives or limbs. It is quite commonly used but not an advisable method to gauge economic losses because not everyone gets themselves insured by insurance companies. The poor section of the society that is often the most affected by road accidents does not actually have insurance and thus this approach has its limitations. Another drawback of this method is that data availability is a challenge as insurance companies do not provide it as an open-source. Eighty percent of accident victims in India are uninsured [12]. The court award approach uses some rationales to value the lost life, and then that amount of money is given to the victim; in the case of a fatal accident, the victims’ relatives can file a case. It is considered an indicator of the cost that society associates with the fatality, or the value that society would have placed on its prevention. Aside from these charges, the cost of real resources is added to this sum to arrive at the cost of an accident. This method is an amalgamation of court awards and other prevalent accident costing methodologies. The verdicts are often used to compensate or calculate the economic losses due to road accidents for future cases of similar intensities. The implicit public sector valuation method tries to put a monetary value on the implicit costs and values of accident prevention in safety legislation or in public sectors. The implicit public sector valuation method employs a set of implicit values to determine the worth of human life.. The willingness to pay method, also known as the value of risk change approach method, seeks to determine how each investment in road safety reduces the likelihood of an accident. It represents society’s readiness to pay to avoid the deaths, injuries, and property loss that result from traffic collisions. Individual and societal willingness to pay can both be calculated by this method. It can be done using the stated preference (SP) and revealed preference (RP) methods. There are several methods under the SP and RP methods, and some of the most well-known ones are listed below. In the Contingent

Meta-Analysis of the Methodologies Used for Road Accident Costing …

683

Fig. 9 Components of road accident cost (HC method) as per miller

valuation method or bidding game survey, respondents are asked their willingness to pay for a certain public good in a given situation. He revealed preference technique attempts to analyse the individual’s or society’s behaviour towards risk and includes a set of questions to gauge the same. Hedonic pricing is used to increase or decrease in benefits of road users, by asking them a bundle of questions. This bundle can have various possibilities of choices. In standard gamble, respondents are given a scenario that which their involvement in a hypothetical road crash results in them having a choice of treatment based on their risk for their outcome. Wage differential method tries to look into how people behave when provided with extra money in exchange for a riskier job. Consumer behaviour tries to look into how the user decides to buy or not to buy based on the price, i.e., a similar technique is said to assess the road safety and price attached to it (Table 1). Table 2 shows a matrix between the various methodologies of road accident calculation and the parameters that are considered for accident costing in the various literatures, the green colour shows the commonly used parameters, the yellow colour is designated for the lesser-used parameters, while the red colour symbolizes the parameters that are rarely considered.

5 Compensation Once a road accident has taken place, it is important to look into the compensation aspects, in order to stabilize the victim, and in case of a fatal accident a “fair Compensation” needs to be calculated in order to stabilize the affected family members of the diseased. As discussed earlier, road accidents may result in acute cognitive disorders in addition to emotional distress and often cause substantial problems in everyday life. In India, the problem is even more grievous as the whole family gets involved when a road accident is recorded, which completely destabilizes the whole family

Advantage (s)

Easiest and most widely used method to be used More suitable when the priority is of national welfare and macroeconomics is to be considered

Based on the valuation by the insurance company and insurer, the victim gets compensation. Generally, the quickest method for receiving compensation

Method

Human capital approach

Life insurance approach

Not everyone gets themselves insured The poor section of the society, that forms a major portion of the victims, does not get covered, i.e., about 80% [10]

Not an advisable method to gauge grief loss and pain Assumes an arbitrary value of pain and grief Widely criticized by economists after the 1970s for not being consistent with (Cost-benefit analysis) CBA principles

Disadvantage (s)

Insurance policy, FIR, an estimate of vehicle repair, indoor patient documents, other related medical bills, death certificate in case of a fatal accident

Property damage, administrative cost, medical cost, lost output

Data required

Table 1 Comparison table of the various methodologies used across the globe

Model dependent on stochastic and rarely deterministic statistical models

Cost-benefit analysis Multi-criterion decision models

Mathematical model

Victim Insurance firm Police Hospitals

Victim Government body Employer Hospitals

Stakeholder

(continued)

Developed and developing countries

Asian and other developing and under developed countries

Widely used in

684 A. A. Azmi and S. Ram

Focuses on looking into the implicitly placed values on accident prevention

Implicit valuation approach

In the valuation of the The extent of damage, cost of a road accident, maintenance cost, depreciation it has a limited use cost, and service life

Often takes too long to Extent of damage, income of the settle cases diseased, might consider “Future Due to long durations, prospects” the victims/their representatives become disheartened and opt out of court

Does not work on fixed rules or formulas, works on guidelines, which can favour the victim Uses past verdicts to decide the valuation of the road accident cost

Court award approach

Data required

Disadvantage (s)

Advantage (s)

Method

Table 1 (continued)

Simply subtracts the explicit costs from the total cost/revenue and then adds that as cost of implicit component of road accidents

It may or may not be based on some previously discussed mathematical models or simply the amount the plaintiff seeks

Mathematical model

Victim insurance firm

Victim Relatives of victim Police second party Insurance companies Government

Stakeholder

(continued)

European and American countries

Almost every country

Widely used in

Meta-Analysis of the Methodologies Used for Road Accident Costing … 685

Disadvantage (s)

Complex method to be applied. Respondents may mention a price for which they are unwilling to pay

No major disadvantages have been found as of now Very limited studies have been done as it is a fairly new method (2020)

Advantage (s)

The best method to gauge the intangible factors like the value of pain and grief It is the most theoretically sound method for the valuation of the human cost

Method

Willingness to pay approach

Hybrid method Combines two approach methods to get the best of both worlds Eliminates the disadvantages of HC and WTP methods

Table 1 (continued) Mathematical model

Property damage, administrative cost, medical cost, Lost output. Stated and revealed preference method to assess the level of pain and suffering and other intangible parameters

Uses a combination of above methods Multi-criterion decision models and probit models

Damage extent to target group, Probit model, stated and revealed preference Multi-criterion method to assess the level of pain decision models and suffering and other intangible parameters

Data required

Victim government body Employer Hospitals Insurance companies

Victim Employer Government hospital

Stakeholder

Limited studies across the globe mainly from Asian countries

Australia, European and American countries

Widely used in

686 A. A. Azmi and S. Ram

Meta-Analysis of the Methodologies Used for Road Accident Costing …

687

Table 2 A matrix of the methodologies and the parameter used for accident costing

especially if the victim is the sole bread earner. Compensation is often a sum of money that compensates only the direct losses completely neglecting the effects on “opportunities lost”, ambitions, and loss of quality of life [41, 48–52]. Compensation is often a frustrating and lengthy process, and people settle for a derisory amount just because they become fed-up with the process, many a time it requires interrogation and re-recording of the accident which further causes mental torment to the victim or their family members, which often degrades the overall health of the diseased and its dependent family members [52, 53, 54–58]. It’s not easy to put a monetary value on life, it’s hard to capture the dynamic nature of costing and compensation with a single statistical or mathematical model. Kenneth R. Feinberg a compensation expert quotes and emphasizes that it’s difficult and debilitating to assign value to life [59, 60]. It was based on the experience of, the Special master, who himself along with his team assigned a monetary value to the death of each 9/11 victim in the USA. He suggested that the value of life equals victims earning over a work-life plus economic loss due to being unable to work and add to that the value of pain and suffering. The formula on paper seems a bagatelle, but in practice, it isn’t. There isn’t something like “One size, fits all” when it comes to costing and compensation [61–63]. India needs to make its unique model that incorporates suitable parameters and hands out compensation that stabilizes the victims and their family members. It has also been found that the acknowledgment of the pain and suffering that the victim or family members are equally important. The compensation should not be too less that it is unable to support the family members, nor it should be too much that one treats it as a lottery. It has been found that in the case when compensation is too much, people often fake their case in order to get that compensation amount [64]. As in the current scenario if one searches for a road accident compensation mechanism in India then either a court case will open up or the Motor accident claim tribunal (MACT) guidelines will appear, which again are mere guidelines and not fixed formulas or rules so that one can easily find the amount of compensation. Thus, a strong mechanism needs to be developed for accident compensation, especially for developing countries like India.

688

A. A. Azmi and S. Ram

6 Results and Discussion 6.1 Limitations and Gaps in the Previous Studies Currently, an updated version of the study titled, “Evaluation of Road Accident Costs” conducted by TCS in 1999 is being used to often decide compensation. This study has few drawbacks and thus the estimates won’t be accurate. Like life expectancy, medical expenses, and valuing the loss of livability had some discrepancies in them, and thus the estimates are not justifiable. As per [55] still “Very few studies have been carried out in India and it is lacked in area coverage and accuracy” [55]. The court award method takes a lot of time and suggests that valuation should be done keeping in view the “Future prospects” of the diseased. But often the valuation is not accepted by the victim or the family members of the victims as it is believed to be quite low by the litigants. Many a times when the verdict is announced half of the people involved in the case are dead. Limited efforts have been made to account for the differential in wages of the different people in the society. Some other social factors that arise in Indian families like loss of opportunities have not been accounted and they still need to be incorporated in the compensation models. Among many methods, there is a debate about which method is most suitable for India. Monetary calculations of pain and grief are still in question and are rarely addressed. None of the methods so far has scientifically analysed the “lost opportunity costs”. Dependent family members are often ignored while handing out the final compensation.

6.2 Proposed Conceptual Framework As discussed in the previous sections of this paper, there are various methods that are used to estimate the cost related to road accidents that can be further used for estimating compensation to the accident victims. There have been a few limitations to each of these methods. This research proposes a conceptual framework for estimating purely the compensation portion for road accident victims that can be used to evaluate the conceptual compensation amount. As pointed out earlier, this framework is developed keeping in view the Indian conditions into the picture so that a “Fair compensation” should be given in a short span of time. This research proposes a name to this compensation equation, i.e. “Ramadilian Equation”. The flowchart of the conceptual framework has been given in Fig. 10. Ramadilian equation for compensation = Restitution cost (C) + Victims related cost (X) + Property damage cost (Y) + administrative cost (Z) where

Meta-Analysis of the Methodologies Used for Road Accident Costing …

689

Fig. 10 Conceptual flowchart of the proposed framework

Restitution cost (C) = Minimum amount of money that should be given to the victim so that they can get their personal and medical expenses covered (especially recommended for low-income and daily wage earners). Irrespective of their fault as this paper believes, saving a life is the ultimate goal. This amount is currently | 50,000/(At max) for grievous accidents and | 2,00,000 (At max) for fatal hit and run cases as per MoRTH guidelines 2022. This paper also proposes that this amount should be inflated as per the inflation rates and at least the medical expenses should be covered for the victims, even if the expenses go beyond the proposed amount of MoRTH. Components of victim-related cost (X): The components of the variable X have been listed in the above flowchart below along with elucidation of whether they are tangible or not. These are the factors that should be given the maximum weightage as they play the most crucial role in deciding the “Fair compensation”. Property damage cost (Y): This variable portion of the equation deals with the costs related to public or private property damages in case of loss of public property, that amount shall be deducted from the compensation amount. Administrative cost (Z): This variable cost deals with the costs incurred during the administrative formalities, that includes amount spend on police investigation, visits to courts, emergency vehicle, etc.

690

A. A. Azmi and S. Ram

7 Conclusion Road accidents are something that has become completely unavoidable. They come along with a lot of social costs attached to them, there are various methods that are used across the globe to measure the economic losses due to road accidents. It has been pointed out that the Human capital costs method and the Willingness to pay method are the most relevant methods when road accident costing is in question. In India, the majority of the studies have used the HC method while limited studies are found using the WTP method. However, after going through all the literature, a fairly new method called the “Hybrid method” seems to be best for a country like India, where we have to look into pecuniary and nonpecuniary aspects. Extending it further this paper proposes Ramadilian equation model for calculating the compensation amount. Compensation needs to be calculated precisely in a scientific manner. Thus, the majority of road accidents should be investigated and compensated accordingly. Compensation can be paid directly to the victim in case of a major and serious accident or it can be provided in the form of credits and those credits can be used for education and medical bills of the victim and their family members. In case of a fatal accident, the immediate family members should get compensation. The maximum time duration for the compensation amount should be in two parts, one as immediate relief which should not take more than 48 h. And the second part should be final compensation that should be handed over in maximum of 2 months. As the compensation has lots of variables the amount of compensation needs to be calculated accordingly. The Ramadilian equation tries to keep all these aspects into the picture and hand out “Fair compensation” in a time- bound manner.

References 1. Sun S (2021) Statista—number of vehicles in operation across India from financial year 1951 to 2019. Transportation logistics ‘vehicles & road traffic (Cross reff) 2. Global status report on road safety (2018) World health organization. France (Cross reff) 3. Verma A, Velumurugan S, Chakrabarty N, Srini S (2011) Recommendations for driver licensing and traffic law enforcement in india aiming to improve road safety. Curr Sci 100(09) (Cross reff) 4. Arbous AG, Kerrich JE (1951) Accident statistics and the concept of accident proneness. J Biom Society 7:340–342 (Cross reff) 5. Deleon MR, Cal PC, Sigua RG (2005) Estimation of socio-economic cost of road accidents in metro manila. J East Asia Soc Transp Stud 6:3183–3198 (Cross reff) 6. Road accidents in India (2019) Delhi: ministry of road transport and Highway, India (Cross reff) 7. Sundar S, Ghate AT (2013) Accidents and road safety: Not high on the government’s agenda. Econ Polit Wkly 48(48):77–83 (Cross reff) 8. Shami S (2005) Road traffic safety: cost of government neglect. Econ Polit Wkly 40(16):1598– 1602 (Cross reff) 9. Das S, Maurya AK (2017) Modelling of motorised twowheelers: a review of the literature. Transp Rev. SSN: 0144–1647 (Print) 1464–5327 (Online) (Cross reff)

Meta-Analysis of the Methodologies Used for Road Accident Costing …

691

10. Singh SK (2017) Road traffic accidents in india: issues and challenges. In: World conference on transport research—WCTR 2016. Shanghai, pp 4708–4719 (Cross reff) 11. Richmund M (2005) Estimation of socio-economic cost of road accidents in metro Manila. J East Asia Soc Transp Stud 3183–3198 (Cross reff) 12. Rajasekaran RB, Rajasekaran S, Vaishya R (2020) The role of social advo- cacy in reducing road traffic accidents in India. J Clin Orthop Trauma 2–3 (Cross reff) 13. Aziz S, Ram S (2021) A review of the international practices for identification of accident black spots and its application in India’s context. Indian Road Congress-Indian High-Ways (Cross reff) 14. Kenney KS, Fanciullo LM, Automobile air bags: Friend or foe? A case of air bag- associated ocular trauma and a related literature review. Optom J Am Optom Assoc 76(7):382–386 (Cross reff) 15. Sugiyanto G (2011) The cost of traffic accident and equivalent accident number in developing countries (case study in Indonesia). ARPN J Eng Appl Sci 12(2) 16. Woyessa AH, Hey WD, Hiko N, Moti BK (2020) Patterns of road traffic accident, nature of related injuries, and post-crash outcome determinants in western Ethiopia—a hospital-based study. Afr J Emerg Med (Cross reff) 17. Dorota M (2008) Social and economic cost of raod accidents in Europe (Cross reff) 18. Sahu PK (2012) Social loss estimation due to road costing NICMAR— J Constr Manag (Cross reff) 19. Kazmia JH, Zubair S (2013, 2014) Estimation of vehicle damage cost involved in Road traffic accidents in Karachi, Pakistan: a geospatial perspective. In: Procedia engineering fourth international symposium on infrastructure engineering in developing countries, IEDC (Cross reff) 20. Osorio CG, Pedraza C (2020) Modern data sources and techniques for analysis. J Traffic Transp Eng 7(4):432–446 (Cross reff) 21. de Guglielmo ML, Caliendo C (2012) Accident rates in road tunnels and social cost evaluation. Procedia—Soc Behav Sci 53(53):166–177 (Cross reff) 22. Chaudhari A, Gore N, Arkatkar S, Joshi G, Pulugurtha S, Exploring pedestrian surrogate safety measures by road geometry at midblock crosswalks: a perspective under mixed traffic conditions. IATSS Res 45(20202021):10–87 (Cross reff) 23. Goyani J, Pawar N, Gore N, Jain M, Arkatkar S (2019) Investigation of traffic conflicts at unsignalized intersection for reckoning crash probability under mixed traffic conditions. J East Asia Soc Transp Stud 13 (Cross reff) 24. Raghuram Kadali B, Vedagiri P (2020) Evaluation of pedestrian crossing speed change patterns at unprotected mid-block crosswalks in India. J Traffic Transp Eng (English edition) 7(6):832 e842 (Cross reff) 25. Chouhan SS, Kathuria A, Sekhar CR (2021) Examining risky riding behavior in India using Motorcycle rider behavior questionnaire. Accid Anal Prev 160:106312 (Cross reff) 26. Paul M, Ghosh I (2021) Development of conflict severity index for safety evaluation of severe crash types at unsignalized intersections under mixed traffic. Saf Sci 144:105432 (Cross reff) 27. Sugiyanto G (2017) The cost of traffic accident and equivalent accident number in developing countries (case study in indonesia). ARPN J Eng Appld Sci (Cross reff) 28. Aziz S, Ram S (2022) A meta-analysis of the methodologies practiced worldwide for the identification of road accidents black spots. Transp Res Procedia 62:790–797 (Cross reff) 29. Bhavan T (2019) The economic impact of road accidents: the case of Sri Lanka. South Asia Econ J 20(1):124–137 (Cross reff) 30. Atubi AO, Gbadamosi KT (2015) Global positioning and socio-economic impact of road traffic accidents in Nigeria: matters arising. Am Int J Contemp Res 5(5):136–146 31. Hu W, Dong Q, Dong C, Yang J (2018) Access to trauma centers for road crashes in the United States. J Saf Res (Cross reff) 32. Rosyidi RM, Priyanto B, Sari SF, Anggraini MA, Kamil M, Wardhana DPW (2020) Subdural drainage of liquor cer- ebrospinal and early tracheostomy: alternative management of severe traumatic brain injury with minimal lesion in limited facility and rural area. Interdiscip Neurosurg 19:100614 (Cross reff)

692

A. A. Azmi and S. Ram

33. Chantith C, Permpoonwiwat CK, Hamaide B (2021) Measure of productivity loss due to road traffic accidents in Thailand. IATSS Res 45(1):131–136S (Cross reff) 34. Reddy GMM, Negandhi H, Singh D, Singh AJ (2009) Extent and determinants of cost of road traffic injuries in an Indian city. 63(12):549–556 (Cross reff) 35. Parkinson F, Kent SJW, Aldous C, Oosthuizen G, Clarke D (2014) The hospital cost of road traffic accidents at a South African regional trauma centre: A micro-costing study. Injury, Int J Care Injured 45(1):342–345 (Cross reff) 36. Parmar J, Das P, Azad F, Dave S, Kumar R (2020) Evaluation of parking characteristics: a case study of Delhi, world conference on transport research—WCTR 2019, Mumbai. Transp Res Procedia 48:2744–2756 (Cross reff) 37. Neelima C, Gupta K, Kumar R, Gautam SP (2014) Analysis of driver behaviour and crash characteristics during adverse weather conditions. Project of CSIR-CRRI (In-House Category) (Cross reff) 38. Dorota Ma´sniak (2008) Social and economic costs of road accidents in Europe (Cross reff) 39. Chinnappa D (2021) Impact of socio-economic profiles on public health crisis of road traffic accidents: a qualitative study from South India. Clinical Epidemiology and Global Health (Cross reff) 40. Road accidents in India (2019) Ministry of road transport and highways, Delhi (Cross reff) 41. Paul SB (2019) Road accidents costing in India. IIT Delhi (TRIPP) and DIMTS (Cross reff) 42. Hills PJ, Jones-Lee MW (1981) The costs of traffic accidents and evaluation of acci- dent prevention in developing countries. In: PTRC summer annual meeting. University of Warwick (Cross reff) 43. Hauer E (1994) Can one estimate the value of life or is it better to be dead than stuck in traffic? Transp Res Ser A 28:109–118 (Cross reff) 44. Miller TR (2000) Assessing the burden of injury: progress and pitfalls. Injury Prevention and Control Taylor and Francis, London (Cross reff) 45. Asian Development Bank.report (Cross reff) 46. World bank group (2021) Socio-economic costs and human impacts of road accidents in Azerbaijan (Cross reff) 47. Viscusi WK (2000) Misuse and proper use of hedonic values of life in legal contexts. J Forensic Econ 13:111–125 (Cross reff) 48. Risbey T (2008) Road crash cost estimation: a proposal incorporating a decade of conceptual and empirical developments, department of transport and regional services (Cross reff) 49. Wijnen W (2021) Socio-economic costs of road crashes in middle-income countries: Applying a hybrid approach to Kazakhstan. Int Assoc Traffic Saf Sci (Cross reff) 50. Giles M (2003) The cost of road crashes: a comparison of methods and recent australian estimates author(s): margaret giles. J Transp Econ Policy (Cross reff) 51. Osterhaudt M (2002) Analysis of accident cost and comparison with available research. Rochester Institute of Technology RIT Scholar Works (Cross reff) 52. Mayou R (1995) Medico-legal aspects of road traffic accidents. J Psychosom Res 39(6):789– 798 (Cross reff) 53. Chen S, Kuhn M, Prettner K, Bloom DE (2019) The global macroeconomic burden of road injuries: estimates and projections for 166 countries. Lancet Planet Health (Cross reff) 54. Balakrishnan S, Karuppanagounder K (2020) Willingness to pay to reduce traffic risk in India. IATSS Res (Cross reff) 55. Nachimuthu K, Partheeban P (2013) Economic analysis of road accident cost for chennai city, india. Em Int 32(4):893–898. issn 0257-8050 (Cross reff) 56. Balakrishnan S, Karuppanagounder K (2020) Estimating the cost of two wheeler road accident injuries in India using the willingness to pay method. Aust J Civ Eng (Cross reff) 57. TRL (2004) The involvement and impcat of road crashes on the poor: Bangladesh and India case studies : A project report by TRL. Transp Res Lab Lond (Cross reff) 58. Elbers NA, Hulstad L, Cuijpers P, Akkermans AJ, Bruinvels DJ (2013) Do compensation processes impair mental health? A meta-analysi. Injury 44(5):674–683 (Cross reff)

Meta-Analysis of the Methodologies Used for Road Accident Costing …

693

59. Melhuish CM (2005) Technical assistance for socioeconomic impact of road crashes (Fi-nanced by the Poverty Reduction Cooperation Fund). Asian Development Bank (Cross reff) 60. Feinberg KR (2006) What is life worth? The unprecedented efforts to compensate the victims of 9/11. Public affairs (US) (Cross reff) 61. Feinberg KR (2012) Who gets what: fair compensation after tragedy and financial up-heaval. Public affairs (US) (Cross reff) 62. Rune E (2018) The value of life: the rise and fall of a scientific research program. Cambridge Scholars Publishing, UK (Cross reff) 63. Graya SE, Colliea A (2022) Work absence following road traffic crash in Victoria, Australia: a population-based study. Injury, Int J Care Injured (Cross reff) 64. Thierauf A, Pollak S, Große Perdekamp M (2009) Simulation of hit-and-run road accidents. Forensic Sci Int Suppl Ser 1(1) (Cross reff) 65. Road accidents in India 2019, Government of India, MORTH, Transport research wing (Cross reff)

Investigating Pedestrian Crash Risk at Unsignalized Midblock Crosswalks on Arterial Road Shubham Thapliyal, Heikham Pritam Singh, and RB. Sharmila

Abstract The increase in pedestrian deaths in road accidents across the developing countries like India is a growing issue of concern. Statistical data indicate that most pedestrian fatalities occurred at midblock crosswalk locations. Pedestrian crossing behaviour data were collected using videography at unsignalized midblock crossing on arterial road. Effect of waiting time, accepted gap, vehicle speed and vehicle type will be considered. Age group, gender, luggage in hand, glancing before walking, all these variables will also be considered in modelling. Reduction in speed of vehicle will also be analysed. The present study will analyse safety margin of pedestrianvehicle conflict under mixed traffic condition at uncontrolled midblock crossing on arterial road. The study will also use both multiple linear regression and the Machine Learning algorithms to model safety margin and assess significant factors affecting it. Rolling behaviour’s effect on pedestrian safety margin will also be analysed. Keywords Midblock crosswalks · Spatial gap acceptance · Safety margin · Vehicle type · Machine learning algorithms · Multiple linear regression

1 Introduction Pedestrian deaths in India have gone up from 13,894 in 2015 to 23,483 in 2020, as per the Union Ministry of Road Transport and Highways. The increase in pedestrian deaths in road accidents across the developing countries like India is a growing S. Thapliyal · H. P. Singh Transportation Systems Engineering, Department of Civil Engineering, Indian Institute of Technology Guwahati, Guwahati 781039, India e-mail: [email protected] H. P. Singh e-mail: [email protected] RB. Sharmila (B) Transportation Engineering, Department of Civil Engineering, Indian Institute of Technology Guwahati, Guwahati 781039, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_55

695

696

S. Thapliyal et al.

issue of concern. It indicates the worrying trend of planning and development of road infrastructure tilting in favour of cars and heavy vehicles. There is a need of giving more attention towards vulnerable road users like pedestrians. Road crossing is considered as the most dangerous location for pedestrians. Main locations for pedestrian crossing are: the midblock crosswalks and intersection (signalized or unsignalized). Statistical data indicate that most pedestrian fatalities occurred at midblock crosswalk locations. Traffic control devices (pavement marking, traffic signals and pedestrian sign board) are used to increase safety of pedestrian at midblock crosswalks. Pedestrian refuge island is made to reduce crossing distance and divide crossing in two or more phases. Frequency and severity of pedestrian accidents are affected by location of crosswalks. Pedestrian demographic factors have an impact on crash outcomes. distraction due to mobile use, luggage in hand, crossing in pair (holding others hand), listening music also affect pedestrian behaviour while crossing. Pedestrian has to do following tasks while crossing road: walking, glancing towards approaching road, maintaining appropriate speed, judging available gap (accept or reject), and negotiating all other distractions. Previous research has shown that some pedestrians who cross road while multitasking fail to display cautionary behaviours (e.g., looking for oncoming traffic, walking inside crosswalk boundary). Vehicle speed is considered as an important risk factor in road accidents. Traffic conditions such as traffic density, speed limits, cross-street activity, and the presence of speedometer influence the speed compliance behaviour of drivers. An arterial road is a high-capacity urban road that sits below freeways/motorways on the road hierarchy in terms of traffic flow and speed. Safety Margin (SM) is defined as the difference between the time a pedestrian crossed the conflict point and the time the next vehicle arrived at the same conflict point [1]. The objective of the present work is to analyse safety margin of pedestrian midblock crossing on arterial road.

2 Literature Review This study investigates pedestrian crash risk with vehicles at unsignalized midblock crosswalks, where vehicles and pedestrians share section of roadway with or without conflicts in a traffic stream. The vehicle can slow down and let pedestrian cross or can accelerate in case of aggressive driving. Pedestrian gap acceptance and road crossing behaviour has been widely studied in the case of signalized intersections [2–4], unsignalized intersections [5, 6], uncontrolled midblock crosswalks [7, 8], and other areas. In the present study, our focus will be on uncontrolled midblock crosswalks on arterial road.

Investigating Pedestrian Crash Risk at Unsignalized Midblock …

697

2.1 Pedestrians’ Gap Acceptance Behaviour Pedestrians’ gap acceptance behaviour at uncontrolled mid-block crosswalks is a crucial aspect of pedestrian safety. The critical gap, which represents the threshold value of the gap acceptance/rejection, plays a significant role in determining whether a pedestrian will safely cross the road. While the arrival time of vehicles has been considered in previous studies, recent research indicates that gap acceptance depends on the distance between the pedestrian and the approaching vehicle [9]. Studies have identified several factors that influence pedestrians’ gap acceptance behaviour, including gap size, crossing distance, pedestrian speed, and vehicle speed [10–13]. Vehicle type has also been found to impact pedestrian’s gap acceptance behaviour [11]. Researchers have observed that the probability of pedestrian platooning increases with the size of the platoon and the traffic volume [8]. Additionally, the frequency of conflicts is likely to increase with an increase in vehicle speed [14]. Pedestrian waiting group composition based on gender, age, disability status, and luggage-carrying capacity has been found to influence gap acceptance behaviour [15]. Safety features such as pedestrian refuge islands, safety guardrails, and road markings have been found to improve the level of service for pedestrian facilities. On urban multilane roads, pedestrians may choose the rolling gap to cross the road, which can be convenient but risky, especially if traffic speed is high [16]. To investigate the factors that contribute to pedestrian-vehicle conflicts, researchers have proposed ordered probit models that consider factors such as traffic volume, vehicle speed, pedestrian crossing behaviour, and pedestrian refuge [14]. Pedestrian safety margin, defined as the time difference between the time gap of an approaching vehicle and the pedestrian’s crossing time, has been used to analyse the severity of pedestrian-vehicle conflicts at unsignalized midblock crosswalks under mixed traffic conditions [17].

2.2 Research Objective Following are the objectives of the present study: • To analyse safety margin of pedestrian-vehicle interaction under mixed traffic condition at uncontrolled midblock crossing on arterial road. • To use multiple linear regression and the Machine Learning algorithms like Decision Tree and Random Forest to classify risk of vehicle-pedestrian interaction based on safety margin values and to check the significant factors affecting risk of vehicle-pedestrian interaction. Categorizes them into three categories, such as low, medium, and high risky interaction.

698

S. Thapliyal et al.

3 Data Collection Video camera was installed to collect pedestrian crossing behaviour data at unsignalized midblock crossing on arterial road (G.S. road) in the city of Guwahati, located in the northeastern part of India. The video data were collected in morning peak hours from 8:00 a.m. to 10:00 a.m. and evening peak hours from 3:00 p.m. to 5:00 p.m. on weekdays. For data collection, one high-definition recording camera was used with recording speed of 25 frames per second. Road geometry was four-lane divided with a 1.95 m wide median, with 7 m width of road on one side of median and 6.7 m width of road on another side of median. The posted speed limit was 30 km/ h on selected road stretch. Pedestrian warning signs were present on site. Raised table top crossing was not found on site. Figure 1 shows Google Earth image of the site selected and data collection setup. Camera setup was such that it could cover pedestrian crossing midblock properly and at least 100 m of approaching traffic from Guwahati to Shilong side. In Fig. 2, road geometry is shown. Light poles are taken as stationary points A, B, and C. Fourth stationary point is taken at the stop line before midblock crosswalk. Distance was measured between these stationary points on ground with the help of distance measuring wheel device. Distance between A & B was 33 m, between B & C was 30 m and between C & D was 20 m. Median width was 1.95 at the midblock crosswalk, which is slightly less than recommended width of median (2 m).

Fig. 1 Data collection setup and google earth image of site

Investigating Pedestrian Crash Risk at Unsignalized Midblock …

699

Fig. 2 Road geometry with stationary points and trap lines on pavement

4 Data Extraction Pedestrian and traffic data were extracted from video recordings manually. Time was noted down when pedestrian: 1. 2. 3. 4. 5.

Enters at the start point of midblock crosswalk. Starts phase 1 journey. Reaches median. Starts phase 2 journey. Reaches other side of midblock crosswalk. The following details were noted down for every pedestrian by visual inspection:

1. 2. 3. 4. 5. 6. 7. 8.

Age category. Gender category. Is he/she glancing towards approaching traffic before crossing? Is he/she crossing street in group or alone? Luggage in hand (hands not free)? Jaywalking (crossing street carelessly or in illegal manner)? Type of accepted vehicle gap (near lane or far lane). Is pedestrian using mobile phone while crossing?

700

S. Thapliyal et al.

5 Data Preprocessing In raw extracted data, there were both categorical and numerical variables. Table 1 shows extracted categorical variables for the modelling while Table 2 shows extracted numerical variables for the study. In Table 2, Speed AD is average speed of vehicle between points A and D. In the same manner, Speed AB will be average speed of vehicle between A and B, Speed CD will be average speed of vehicle between C and D. Difference of Speed AB and Speed CD is also taken as a parameter Speed AB–CD, which shows reduction in speed of vehicles while approaching pedestrian crosswalk. Crossing speed at phase 1 is also taken for modelling. Table 1 Categorical variables used in modelling Serial number

Variables

Categories

1

Age group

1. Child 2. Young 3. Old

2

Gender

1. Male 2. Female

3

Grouping

1. Yes 2. No

4

Glancing before walking

1. Yes 2. No

5

Luggage in hand

1. Yes 2. No

6

Vehicle type

1. 2. 3. 4. 5.

7

Accepted gap type

1. Near 2. Far

8

Rolling behaviour

1. Yes 2. No

Table 2 Numerical variables used in modelling

Serial number

TW (two-wheeler) A (Auto) C (Car) LCV (Light commercial vehicle) HV (Heavy vehicle)

Variables

Unit of parameter

1

Wait at start

s

2

Accepted spatial gap

m

3

Speed AD

m/s

4

Speed AB–CD

m/s

5

Crossing speed phase 1

m/s

6

Safety margin

s

Investigating Pedestrian Crash Risk at Unsignalized Midblock … Table 3 Accident risk based on safety margin value

Safety margin range (sec)

701

Accident risk

0–2.5

High risk

2.5–6

Medium risk

≥6

Low risk

Fig. 3 Histogram of pedestrian safety margin values

Figure 3 shows histogram of pedestrian safety margin values. Following Table 3 shows the corresponding accident risk for pedestrian based on safety margin values. Risk of accident increases with decrease in safety margin value, so higher the safety margin value safer the pedestrian. In this study, a multiple linear regression and different machine learning algorithms are used to classify accident risk of pedestrian and their performance is compared.

6 Preliminary Data Analysis and Results There were 754 pedestrian data points in the study, out of which 33% were female and 67% were male pedestrians. Considering the age categories, 11.3% pedestrian were children, while only 3.8% were old pedestrians and others were young pedestrians. Out of all pedestrians, 13.3% were carrying luggage. Table 4 shows mean, standard deviation, minimum and maximum value for safety margin, crossing speed, vehicle speed, change in vehicle speed, accepted gap and waiting time. In data analyses, average safety margin was found to be 3.03 s. Average crossing speed of pedestrian was 1.163 m/s. For the leading vehicle, average reduction in speed while approaching to midblock crosswalk was 0.279 m/s (1 km/h). So, it was concluded that vehicles slow down very slightly while approaching to midblock

702

S. Thapliyal et al.

crosswalk. Average waiting time was 9.3 s for the pedestrian during morning and evening peak hours in our study. When leading vehicle was in far lane, average accepted gap for the pedestrian was 38.25 m and when it was in near lane, average accepted gap was 27.10 m. It concludes that pedestrian has tendency to accept bigger gap when leading vehicle is in far lane. We can also say that pedestrian gap acceptance behaviour changes as per the position of leading vehicle in different lanes. Average safety margin value when the approaching vehicle speed was 1 [14]. Other than B–C ratio, other methods like net present value (NPV) method and economic internal rate of return (EIRR) are also deployed, and all the three methods are based on the discounted cash flow (DCF) technique of discounting all future costs and benefits

Economic Benefit Assessment of Black Spot Improvements

719

to a common year. The three methods used for economic evaluation are discussed below in brief [14].

8.1 Benefit–Cost Ratio Method There are a number of variations of this method, but a simple procedure is to discount all costs and benefits to their present worth and calculate the ratio of the benefits to costs. Negative flows are considered as costs whereas positive flows as benefits [14]. Benefit − cost ratio =

Total benefits over the analysis years discounted to the reference year Total cost over the analysis years discounted to the reference year

8.2 Net Present Value Method In this method, the stream of costs/benefits associated with the project over an extended period of time is calculated and is discounted at a selected discounted rate to give the present value. Benefits are treated as positive and costs as negative and the summation give the Net Present Value (NPV) [14]. NPV0 = (B0 − C0) +

Bn − Cn B1 − C1 + ··· + 1 (i + n)n (i + 1)

8.3 Economic Internal Rate of Return Economic internal rate of return (EIRR) is the discount rate that makes the discounted future benefits equal to the initial outlay. In other words, it is the discount rate that makes the stream of cash flows to zero. The solution to the equation given below can be done by trial and error. However, the task of computing EIRR is rendered very simple nowadays due to the availability of this function as an inbuilt one in many software. If the EIRR calculated from the above formula is greater than the rate of interest obtainable by investing the capital in the open market, the scheme is considered acceptable [14]. A summary of the deduced results for the four black spots is presented in Table 4.

720

M. Balaji et al.

Table 4 Average unit costs of crashes for each crash severity Type of severity

Average unit crash cost per victim (INR)

Fatality cost

91,16,363

Grievous injury cost

3,64,398

Minor injury and property damage cost

83,201

9 Conclusions This study focused on analysing four severe black spots in the city of Nagpur and provided an understanding of the economic benefit assessment of the proposed countermeasures. Some of the salient findings derived are as follows: • The proposed countermeasures were found to be cost-effective for the four black spots conforming to IRC:131 [3]. It is estimated that about 60–66% reduction in the overall road crashes coupled with 40% reduction in fatalities if the countermeasures are applied assuming a similar rate of road crashes in the next 5-year period on the road network of Nagpur as black spot improvements falls under cost-effective improvements. • The total cost savings for each black spot are summarized in Table 5. On average, there is a cost-saving of about Rs. 89.15 lakhs in the midblock locations and Rs. 1.26 crores in the intersections after the implementation of 1 year of proposed countermeasures. • The Economic Internal Rate of Return (EIRR) was found to be ranging between 54% and 63% through the analysis period of 5 years, which can be considered to be a significant return on investment (ROI). Even the First Year Rate of Return Table 5 Summary of benefit–cost evaluation Name of the black spot

Benefit–cost ratio

First-year cost savings (INR)

First-year rate of return (%)

Net present value

Economic internal rate of return

Chikli square—option 1

9.82

1,18,02,695

2.71

4,21,19,210

63.53

Chikli square—option 2

5.16

1,24,36,765

1.42

4,04,63,423

54.72

10.01

85,32,006

2.76

3,05,13,013

63.71

Telephone exchange to C.A.

6.22

92,96,482

1.72

3,09,96,116

57.89

Jhansi Rani square 1

5.46

1,33,88,615

1.51

4,34,54,691

55.76

Prakash High school

Economic Benefit Assessment of Black Spot Improvements

721

(FYRR) was estimated to be ranging between 1.4 and 2.76%, which shows that there is bound to be an immediate ROI in all four black spot locations. • At the same time, though the implementation of black spot interventions is an effective tool in treating the affected road stretches/locations, it is to be borne in mind that it is one of the forms of interventions to address road safety in the road network of any city.

10 Future Scope The limitations and the future scope of this study are as follows: • More robust form of data collection is necessary to account for unreported and minor injury crashes. • The effectiveness of the countermeasures needs to be studied extensively in the Indian road conditions. • Countermeasures need to be studied for its medium to long-term effectiveness by using reliable projections of the traffic and crash data through studying the trends in road crashes.

References 1. Road accidents in India (2020) GoI. Published by Ministry of Road Transport & Highways, New Delhi 2. Intelligent Solutions for Road Safety through Technology & Engineering, Interim Report (2022) A report submitted to Nagpur Municipal Corporation 3. IRC: 131-2022. Guidelines for identifying and treating black spots. Indian Roads Congress 4. Duduta N, Adriazola-Steil C, Wass C, Hidalgo D, Lindau LA, John VS (2015) Traffic safety on bus priority systems: recommendations for integrating safety into the planning, design, and operation of major bus routes. EMBARQ/World Bank Group, Washington DC 5. FHWA Safety (2013) Traffic calming countermeasures library. Safer Journey. http://safety. fhwa.dot.gov/saferjourney/library/ 6. Rosen E, Sander U (2009) Pedestrian fatality risk as a function of car impact speed. Accid Anal Prev 41:536–542 7. Elvik R, Hoye A, Vaa T (2009) The handbook of road safety measures. Emerald Group Publishing, Bingley 8. Sul J (2014) Korea’s 95% reduction in child traffic fatalities: policies and achievements. The Korean Transport Institute (KOTI), Seoul 9. Road crash analysis and evaluation of mitigation measures using iMAAP software of an interurban corridor (2020) by Angel Maria Mathew 10. Development of Model Road Stretches of NHAI (under Policy Guidelines No 18.44/2020) 11. Real GDP growth percentage, The World Bank. https://data.worldbank.org/indicator/NY.GDP. DEFL.KD.ZG?locations=IN 12. IRC: 35-2015. Code of practice for road markings (Second Revision). Indian Roads Congress 13. IRC: 67-2022. Code of practice for road signs (Third revision). Indian Roads Congress 14. IRC: SP: 30-2019. Manual on economic evaluation of highway projects in India (Third revision). Indian Roads Congress

722

M. Balaji et al.

15. Welle B, Li W, Adriazola-Steil C, King R, Obelheiro M, Sarmiento C, Liu Q (2015) Cities safer by design. Urban design recommendations for healthier cities, fewer traffic fatalities. World Resources Institute 16. Study on Socio—Economic Cost of Road Accidents in India. A study by DIMTS, in association with TRIPP, IIT-Delhi. Published by Ministry of Road Transport & Highways 17. World Bank (2017) The high toll of traffic injuries: unacceptable and preventable. © World Bank

Emerging Technology, Logistics and Sustainability

Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway V. A. Ajay Swaroog, Sheela Alex , and Padmakumar Radhakrishnan

Abstract An autonomous vehicle (AV) or driverless car ensures safety using advanced sensor and positioning technologies with little or no human input. With numerous emerging technologies, it is important to know the potential capacity effects of AVs to aid our decision-making process with future investments. As AVs are not widely used especially in developing countries like India, it is difficult to study the behaviour of AVs and their interactions with other vehicles in the field. Hence, the mid-block section of a two-lane highway incorporating AVs is simulated using VISSM in this study to analyse how AVs impact the capacity of the highway. Passenger cars are replaced by autonomous cars in the model and the model is calibrated using travel time as measure, and the change in the capacity for various penetration rates of AV is estimated. The study gave promising outputs in terms of capacity enhancement and found that the introduction of the AVs can enable better utilization of the available road space. Keywords Autonomous vehicle · Capacity · Two lane highway · VISSIM

1 Introduction An autonomous vehicle (AV), also known as a self-driving car or driverless car, is a vehicle that can detect its environment and move safely with little or no human input. AVs use various sensors such as radar, lidar, sonar, GPS, odometry and inertial

V. A. Ajay Swaroog (B) · S. Alex · P. Radhakrishnan Department of Civil Engineering, College of Engineering Trivandrum, Thiruvananthapuram, Kerala, India e-mail: [email protected] S. Alex e-mail: [email protected] P. Radhakrishnan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_57

725

726

V. A. Ajay Swaroog et al.

measurement units to perceive their surroundings. In 2014, automotive standardization body SAE International published a classification system of AVs with six levels ranging from fully manual (Level 0) to fully automated systems (Level 5) [1]. At present, Level 3 AVs are available for consumers to purchase. Today AVs are being introduced in the market and have the potential to mitigate some of the negative impacts of transportation [2]. They can improve highway safety and efficiency by decreasing driver inputs, reducing driver workloads and human error, and providing environmental benefits by reducing emissions and fuel consumption. Many of the benefits of AVs will not be obtained until after full market penetration, which is not expected to happen immediately. It is estimated that autonomous cars will reach 50% global penetration only by the year 2024. With numerous emerging technologies, it is important to understand the potential capacity effects of AVs to aid our decision-making process with future investments. Adebisi et al. [3] conducted simulations on basic freeway, freeway merge, and freeway weaving segments using two major AV applications: cooperative adaptive cruise control and advanced merging. The study discovered that as the market penetration rate of Connected AVs (CAVs) increases, it can significantly improve roadway capacity by up to 35–40% under certain circumstances. This improvement was observed not only on basic freeways but also on merge and weaving segments. Park et al. [4] conducted a study on urban roads to investigate how gradual increments of AV penetration and traffic volume impacted traffic flow and road capacity. The study discovered that as AV penetration increased, traffic flow improved with a reduction of up to 31% in average delay time. Links with three or four lanes had a more significant impact on delay reduction than those with fewer lanes. When AV penetration reached 100%, the road network could accommodate 40% more traffic. Olia et al. [5] discovered that if all vehicles are driven in a cooperative automated manner, it is possible to achieve a maximum lane capacity of 6,450 vph per lane (300% improvement). The study also found that incorporating AVs into the traffic stream did not significantly affect achievable capacity. Zhao et al. [6] built VISSIM simulation models for different sections of freeway under different traffic status scenarios and various levels of autonomous vehicle penetration to quantify the traffic impact under multiple penetrations of autonomous vehicles. The analysis showed that in free flow scenarios, penetration of autonomous vehicles had a slight positive impact on traffic operations for all segments. Ye et al. [7] studied how CAVs affect traffic safety under various penetration rates and discovered that as CAV penetration rate increased, smooth driving increased while velocity difference between vehicles decreased which greatly smoothed out traffic flow. Stopand-go traffic was also greatly eased. Majority of the studies were done on freeways using simulation software and not on Indian traffic condition. In most literature, they arrived at the conclusion that AVs to have a positive impact on traffic and safety. Traffic in India is very different compared to the lane following traffic in developed countries. Different types of vehicles like auto and truck share the same lane as two-wheelers and cars. This brings the need to study the effect of AVs on our existing traffic condition when they are moving along with other vehicles. With numerous emerging technologies, it is

Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway

727

important to know the potential capacity effects of AVs to aid our decision-making process with future investments. In this study, the effect of AVs on the capacity of a two-lane highway is studied. The scope of the study covers developing simulation model of a mid-block section of a four lane divided National Highway and the calibration and validation of the simulation model incorporating AVs using Travel time as the characteristic.

2 Methodology 2.1 Data Collection A two-lane unidirectional highway of the Bypass near Thiruvallam, Thiruvananthapuram is selected for the study. An aerial view of the site is given in Fig. 1. A site visit is done and geometric details of the section are collected. In order to obtain data like volume, traffic composition and so on video data were collected from a foot overbridge near the section. Geometric Details. The study location is a two-lane unidirectional highway. Each lane is 3.5 m. The left shoulder is of 1.5 m width, and the right shoulder is 30 cm wide. Video Data. Video data were collected for 2 h from 3:30 pm to 5:30 pm. Camera and tripod were set on a foot overbridge near the location. It was possible to collect data through a distance of 160 m.

Fig. 1 Aerial view of the NH bypass near Thiruvallam, Thiruvananthapuram

728

V. A. Ajay Swaroog et al.

Fig. 2 Traffic composition

Traffic volume was counted manually and was found to be 1060 veh/h. Traffic composition was also found by manual method. The video was played on a computer and count was taken. The traffic composition is given in Fig. 2. Two-wheelers dominated with 51% of the traffic. It is followed by cars with 37% and three-wheelers with 7%. Mainly three types of cars were seen on the field, hatchback, sedan and jeep or five-seater. Also, the buses were divided into mini bus and normal bus. Heavy vehicles were divided into three categories based on their dimensions. To find average travel time of vehicles, KinoveaTM software was used. Using the KinoveaTM software, grid lines were drawn on the video data from start to finish of the 160 m stretch. Then travel time of each type of vehicle was found using manual method. The average travel time of entire traffic was 9.8 s. The average speed was also found.

2.2 Simulation Model Building Simulation model was built using PTV VISSIM 2021 micro-simulation software and the snapshot of the model is given in Fig. 3. PTV VISSIM is a microscopic multi-modal traffic flow simulation software package developed by PTV Planning Transport Verkehr AG in Karlsruhe, Germany [8]. The geometric details, traffic volume and composition collected from field were input into the Wiedmann 74 model of VISSIM. Autonomous vehicles (Wiedemann 99 model is used) replace passenger cars only, with two types of autonomous cars, namely cautious and aggressive types. AVs were not allowed to overtake on the same lane, and pedestrian movement and interactions are not considered in the model.

Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway

729

Fig. 3 Snapshot from VISSIM simulation

Calibration of Simulation Model. Simulation model built in VISSIM by inputting geometric details, traffic volume and composition will not be an accurate representation of the field condition. In order to achieve that, various lateral and driving behaviour-related parameters need to be adjusted to match the field condition. PTV VISSIM user manual [9] consists of all the different parameters, their typical range and based on that calibration is done. The values of default and calibrated parameters of the model are given in Table 1. To check whether the simulation is an accurate representation of field condition, travel time from field data and simulation is compared. The travel time for different vehicle types for a certain distance in the field is obtained from video data. Travel time for the same length of road is also found from simulation and compared with Table 1 Default and calibrated values of parameters in the model Parameter

Default value

Calibrated value

0

10

Driver behaviour parameters Look ahead distance (minimum) (m) Look back distance Minimum

0

10

Maximum

150

50

Weidmann 74 car following characteristics Average standstill distance (m)

2

2.5

Overtake on same lane

No

Yes

Minimum lateral distance at 0 and 50 km/h (m) Two-wheelers

1, 1

0.50, 0.75

Three-wheelers

1, 1

0.50, 0.75

Car

1, 1

0.50, 1.00

Heavy vehicle

1, 1

0.75, 1.00

LCV

1, 1

0.75, 1.00

Bus

1, 1

0.75, 1.00

730

V. A. Ajay Swaroog et al.

field data and MAPE value is calculated. Before calibration, the MAPE value was 33%, which reduced to 3.18% after calibration. Simulation of AVs. To determine the impact of AVs, simulation is run with increasing penetration of AVs. That is, at first 10% of the cars in the simulation are converted to autonomous cars, then 20% and so on until the entire cars are converted to autonomous cars. Only cars are given autonomous features in simulation since it is autonomous cars that are mainly sold to consumers. Autonomous cars are not yet introduced in India, so it is not possible to model them based on observation from field. So, AVs are simulated in VISSIM based on previous studies. CoExist simulation study [10] provides insight into simulation of AVs in mixed traffic using VISSIM. Different studies have made different adjustments to the default values for modelling AVs. These studies have made some common adjustments also, like AVs keeping a shorter distance with the front vehicle, having faster and smoother reactions and observing more around vehicles. Two types of autonomous cars, cautious and aggressive type were used in the simulation. Cautious AVs are the ones that are initially introduced in the market. They keep more headway and less acceleration values than that are capable by the vehicle so that general public starts trusting AVs. As the AVs become more and more accepted aggressive AVs will be released. The parameters and the values used for aggressive and cautious vehicles are given in Table 2. Table 2 Simulation parameters used for simulation of AVs [11, 12] Parameter

Aggressive

Cautious

CC0 (standstill distance) (m)

1

1.5

CC1 (headway time) (s)

0.6

1.5

CC2 (‘following’ variation) (m)

0

0

CC3 (threshold for entering ‘following’)

−6

−10

CC8 (standstill acceleration)

(m/s2 )

CC9 (acceleration at 80 km/h) (m/s2 )

4

3

2

1.2

Minimum look ahead distance (m)

0

0

Maximum look ahead distance (m)

300

250

Minimum look back distance (m)

0

0

Maximum look back distance (m)

150

150

Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway

731

3 Results 3.1 General The simulation was run with increasing penetration of AVs. The simulation was run for a duration of three hours. The first and last 30 min of the simulation were taken as warm up period and not considered for data collection. The total number of different vehicles in the simulation is obtained as output. From this, volume in veh/h is found and then converted to PCU/h.

3.2 Capacity Determination Simulation runs were used to determine capacity by increasing inflow from nearzero volume to higher volumes during successive simulation runs and calculating throughput (outflow) from simulation output. At lower inflows, outflow volume increases as vehicle input (inflow) increases but when maximum throughput is attained, higher vehicle input will not result in same increment in outflow volume and decrement in outflow despite the increase in vehicle input in consecutive simulations shows that freeway segment reaches its capacity [13]. This method is used to determine the capacity of each simulation run.

3.3 Simulation of Cautious AVs Simulation was run with increasing penetration of cautious AVs and capacity of each simulation run was determined. The capacity at different penetration of AVs is given in Fig. 4. Not much change in capacity was observed at 0–100% penetration of AVs. Only 1% improvement in capacity was observed when penetration was 40%. Maximum improvement was 2% observed at 70% penetration of AVs. Capacity actually reduced by 0.5% at 100% penetration of AVs. This might be due to cautious AVs keeping more headway and driving with less acceleration compared to HDV. These vehicles also follow lane and won’t overtake on the same lane. This might be the reason why improvement in capacity was low even at 100% penetration of cautious AVs.

3.4 Simulation of Aggressive AVs Simulation was run with increasing penetration of aggressive AVs and capacity of each simulation run was determined. The capacity at different penetrations of AVs is

732

V. A. Ajay Swaroog et al.

Fig. 4 Percentage increase in capacity at increasing penetration of AVs for cautious AVs

given in Fig. 5. It was observed that the capacity increases with increasing penetration of AVs. An increase of 10% was observed at 50% penetration of AVs and an increase of 19% was observed at 100% penetration of AVs. Even though AVs were not allowed to overtake on the same lane their presence improved capacity by eliminating human error and keeping lower headway than possible by HDV.

Fig. 5 Capacity at increasing penetration of AVs for aggressive AVs

Impact of Autonomous Vehicles on Capacity of a Two-Lane Highway

733

4 Conclusion In this study, VISSIM is used to determine the impact of AVs on capacity of a two-lane highway. The impact of both cautious and aggressive AVs was determined separately. Cautious AVs did not have much impact on capacity (maximum improvement was 2% at 70% penetration rate); the capacity values were observed to be fluctuating instead. This might be due to the AVs keeping more headway and having comparatively less acceleration, thereby forcing other vehicles to overtake and proceed. On the other hand, aggressive AVs improved capacity by 19% at 100% penetration of AVs. For them, the capacity increased with increasing penetration of AVs. Hence, based on the study, it may be concluded that the presence of AVs would bring a positive impact on our roads. Impact of autonomous vehicles on the behaviour (safety) of other vehicles and pedestrians in the stream is not addressed in the study. Wiedemann 99 model for all vehicles may help in better modelling the microscopic levels of interaction between vehicles and AVs in a heterogeneous traffic flow. Calibration of the model is done at a macro level; comparison of parameters (speed, acceleration, lane changing) at micro-level is essential to accurately model heterogeneous traffic flow incorporating AVs. Further studies are necessary incorporating a blend of cautious and aggressive as well as hybrid AVs in the stream, and analysing the impact of these changes on the capacity, level of service and safety on the highways.

References 1. Synopsys Automotive webpage. https://www.synopsys.com/automotive/autonomous-drivinglevels.html 2. Fagnant DJ, Kockelman K (2015) Preparing a nation for autonomous vehicles: opportunities, barriers and policy recommendations. Transp Res Part A: Policy Pract 77:167–181 3. Adebisi A, Liu Y, Schroeder B, Ma J, Cesme B, Jia A, Morgan A (2020) Developing highway capacity manual capacity adjustment factors for connected and automated traffic on freeway segments. Transp Res Rec 2674(10):401–415 4. Park JE, Byun W, Kim Y, Ahn H, Shin DK (2021) The impact of automated vehicles on traffic flow and road capacity on urban road networks. J. Adv. Transp. 2021 (2021) 5. Olia A, Razavi S, Abdulhai B, Abdelgawad H (2018) Traffic capacity implications of automated vehicles mixed with regular vehicles. J Intell Transp Syst 22(3):244–262 6. Zhao N, Zheng S, Hao S, Li J, Wu K, Jiao C (2020) Traffic impact analysis of typical sections of freeway under multiple penetrations of autonomous vehicles. In: CICTP 2020, pp. 728–738 (2020) 7. Ye L, Yamamoto T (2019) Evaluating the impact of connected and autonomous vehicles on traffic safety. Phys. A 526:121009 8. Wikipedia webpage. http://en.wikipedia.org/wiki/PTV_VISSIM. Accessed 11 July 2022 9. VISSIM User Manual, Version 2020.00-11 10. Sukennik P (2018) PTV Group. Micro-Simulation Guide for Automated Vehicles. COEXIST (h2020-coexist.eu) 11. Aria E, Olstam J, Schwietering C (2016) Investigation of automated vehicle effects on driver’s behavior and traffic performance. Transp. Res. Procedia 15:761–770

734

V. A. Ajay Swaroog et al.

12. Stogios C, Kasraian D, Roorda MJ, Hatzopoulou M (2019) Simulating impacts of automated driving behavior and traffic conditions on vehicle emissions. Transp Res Part D: Transp Environ 76:176–192 13. Beza AD, Zefreh MM, Torok A (2020) Impacts of different types of automated vehicles on traffic flow: an experimental analysis

Quality Assessment of App-Based Bike Taxi Services by Benchmarking and Numerical Rating Approach: Guwahati Lalit Swami , Mokaddes Ali Ahmed, and Suprava Jena

Abstract Shared mobility is an umbrella term that includes various forms of bikesharing, carsharing, carpooling, vanpooling, and on-demand ride services. It also includes feeder transport services like shuttles and paratransit. With the advancements in technology, options for shared mobility are evolving. In Indian cities, appbased mobility services have become an emerging business. App-based bike taxi service is also an innovative shared mobility model that has emerged recently. This study evaluates the quality of app-based bike taxi services in Guwahati by benchmarking and numerical rating approach (NRA). Key Performance Indicators (KPIs) were developed and derived from various studies based on existing scenarios. A questionnaire survey was conducted, and then NRA was used to evaluate the quality. Additionally, this study explores the aspects that make these services more reliable and accessible, causing commuters to switch. The overall level of service (LOS) of app-based bike taxi services is found two, indicating room for improvement in some areas. The study finds that app-based bike taxi services perform better in realtime tracking, affordability, service coverage, comfort, and travel speed. However, the extent of supply of service needs to be improved. By NRA, deficiency from an acceptable level is spotted in the luggage carrying capacity of the service. In addition, app-based bike taxi services provide online ride-booking, door-to-door pickup and drop service, live location tracking, and multiple fare payment options, making it more attractive than other modes. Keywords Shared mobility · Bike taxis · Bike-sharing · App-based bike taxis

L. Swami (B) · M. A. Ahmed · S. Jena National Institute of Technology Silchar, Assam 788010, India e-mail: [email protected] M. A. Ahmed e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024 D. Singh et al. (eds.), Transportation Research, Lecture Notes in Civil Engineering 434, https://doi.org/10.1007/978-981-99-6090-3_58

735

736

L. Swami et al.

1 Introduction Shared mobility is a transportation model that enables users to gain short-term access to transportation modes on an “as-needed” basis. It is an umbrella term that includes various forms of carsharing, bike-sharing, vanpooling, carpooling, and on-demand ride services. Shared mobility can also include transit services like shuttles, paratransit, and private transit services. On-demand ride services based on apps have recently shifted the paradigm of urban transport [1]. These services are provided by transportation network companies like Ola and Uber and are also termed ridesourcing services. Using a smartphone to request trips from possible suppliers in real time, Ridesourcing continuously balances demand and supply. Ridesourcing services are distinguished from traditional taxi services by the use of smartphones and ridematching techniques [1]. As the options for urban mobility are evolving, App-based bike taxi service is also an innovative model that has emerged recently. App-based bike taxi service is a demand-driven system where the booking is made via an Android app, and one bike rider carries one pillion rider. Many ridesourcing platforms like Ola, Uber, and Rapido have started providing rides by motorcycles, scooters, or e-bikes. App-based bike taxi services have emerged as a disruptive mode in Indian cities for transporting and delivering goods. However, they have also generated concerns among many regulators. As city authorities explore policies on these new innovative modes like app-based bike taxi services, independent data on use, service quality, and impacts are highly required, this study aims to fill this research gap and provide a quality evaluation of app-based bike taxi services by benchmarking. The NRA is also used for considering passenger satisfaction and their perception so that this process can be used in any Indian city.

1.1 Background App-based bike taxi services permit users to request a ride through an Android application; this notifies nearby drivers of the passenger’s location. After the ride is accepted, the users can see the driver’s current location and expected arrival time. The smartphone application offers location tracking, making it easier for inexperienced drivers to find their destinations and lowering the likelihood that they would take a wrong turn [2]. Prices respond dynamically to demand, and passengers can pay the fare by UPI, credit card, or cash. At the end of the trip, drivers and passengers rate each other, creating a system of rewards that encourages polite behaviour. Unlike traditional taxis or rickshaws, Ola, Uber, and Rapido drivers lack a commercial vehicle permit, drive their personalized scooter, motorcycle or e-bike, and work part-time. The nomenclature used to describe these services has been highly debated. Other names for app-based taxi services include: “real-time ridesharing,” “ride-hailing,” “bike taxis,” “on-demand rides,” “bike-sharing,” and “ridesorcing” [1]. The name

Quality Assessment of App-Based Bike Taxi Services by Benchmarking …

737

“App-based bike taxi services” is used in this study because it precisely conveys the fundamental technology- an application where a motorcycle or scooter can be hired for a ride. However, definitions are confusing, mainly because these services continue to develop.

1.2 Related Literature Research on new shared mobility models (ridesharing/ridesourcing) is used to shed light on the expected usage, features, and performance of app-based bike taxi services because independent research on App-based bike taxi services is quite limited. City authorities have supported ridesharing for decades because it reduces the vehicle miles travelled. Individually, ridesharing users benefit from reduced commute stress, travel costs, and journey time from high occupancy vehicle lanes [3]. Despite the many advantages, ridesharing has encountered several obstacles, such as a willingness to give up the convenience, comfort, and flexibility of the personal automobile, a need for privacy and personal time, and concerns about personal safety when riding with unknown people [4]. Taxi services cater to a number of markets, including senior citizens, wealthy individuals, and low-income households without cars. Despite having a low modal share, taxis serve an essential need by offering transportation when cars or other public transportation forms are unavailable [5]. Shared taxis may have advantages like higher efficiency, lower passenger costs, and reduced traffic and overall vehicle usage [6]. However, sharing a cab with strangers is often not allowed in some countries like America. In the street hail, customers cannot compare information on fare or quality before selecting a ride, a lack of information is really a problem. A lack of information is an issue since customers cannot compare information on fare or quality before selecting a vehicle. This leads to poor service quality. Due to the low entry barrier in the taxi industry, these markets have over competition. This leads to reckless and aggressive driving, poorly managed vehicles, and congestion [7]. Regulatory responses to these concerns include control over market entry or supply, tariff regulations, and vehicle and driver safety standards. Furthermore, technological advancements raise concerns about how the necessity for regulation may have evolved. Rating systems may be successful in overcoming the lack of information issue. People are now not required to wait on the street or make a phone call to hail a taxi or auto rickshaw. Passenger can now track their own and the driver’s whereabouts in real-time. App-based bike taxi services have traits similar to traditional taxis but also have the ability to use the advantage of technology. App-based bike taxi services possess a challenge for regulators. Addressing these challenges clearly requires more information regarding their quality and use in Indian cities.

738

L. Swami et al.

2 Study Area Guwahati is the largest city in N-E region. It is also the largest industrial, commercial, and educational hub in the region. Furthermore, significant number of students and workers from other northeastern states are used to migrate to the city. Traffic jams have become a regular phenomenon in the city, due to its increasing population growth and limited infrastructure and road development. The public transit system comprises organized and unorganized public transport. The organized system contains city bus services operated and managed by Assam State Transport Corporation. The unorganized public transportation system consists of e-rickshaws, App-based taxis, traditional taxis, and other paratransit services. RAPIDO aggregator launched the app-based bike taxi services in Guwahati in 2018, putting bike-sharing on the online portal to provide users with a smooth and trouble-free ride experience. Ola, Uber, and Rapido are currently major aggregators in Guwahati providing app-based bike taxi services in Guwahati by booking through their online apps.

3 Methodology The strategy adopted for the research to achieve the objectives is displayed in Fig. 1. Seven key performance indicators are considered to evaluate overall performance and identify performance areas that require improvement. Likewise, for NRA, 11 service attributes indicating the service quality of app-based bike taxi services are recognized. Together, the two evaluation techniques show where the app-based bike taxi excels and where it needs to improve.

3.1 KPI for Benchmarking The performance areas have to be recognized to evaluate the performance of appbased bike taxi services in Guwahati by benchmarking. Comfort, mobility, convenience, affordability, and ITS facilities are the performance areas that are essential for the trip makers and urban development authorities, as stated in Table 1.

3.2 Service Attributes for Numerical Rating Approach Eleven service attributes are considered essential to evaluate the service quality from the passenger’s point of view. Service quality is the difference between passenger expectations and perception of app-based bike taxi service. Self-rating questionnaires were used as a data collection method in this study. The attributes considered are given

Quality Assessment of App-Based Bike Taxi Services by Benchmarking …

739

Study area determination for quality assessment of app-based bike taxi services

Numerical rating

Benchmarking

Service attributed based on passenger perspective

Key performance indicator

Detailed analysis of data

Data collection using primary survey and secondary sources

Detailed analysis of data

Evaluating performance area of bike sharing system

Evaluating performance area of bike sharing system

Identifying areas that need improvement

Fig. 1 Flowchart showing the methodology

in Table 5. The rating of the service attributes is based on individual experience and varies between individuals. The relative weights are given among attributes based on preference by passengers. Travel costs reflect the requirements for affordability, and depend on the income of the trip maker. Driver behaviour demonstrates the driver’s attitude, complying with traffic rules, drinking, and driving.

4 Data Collection and Extraction A trip maker survey was performed at major locations in Guwahati by random sampling. About 200 random passengers were asked to grade using a Likert scale ranging from very good (5) to poor (1). To assess the relative weight of various service attributes, passenger attitude survey was performed on a Likert scale. Speed and delay studies were conducted along major routes to find the average travel speed of app-based bike taxi services.

740

L. Swami et al.

Table 1 Key performance indicators for benchmarking Performance measure

Performance indicators Method of KPI determination

Standard’s derived from

Availability

Extent of supply

Supply of vehicle/1000 population

[8]

Service coverage

Distance in route on which service services to the entire urban

[9]

Comfort

Driver behaviour, fare collection system, simplicity to use service, luggage carrying capacity

Index of acceptability

Author

Convenience

Waiting time

Average waiting time from trip maker survey.

Author

Mobility

Travel speed

Travel speed of service along major routes in the city

[8]

Affordability

Affordability

The ratio of income [8] spent on bike taxi trips to their monthly income

ITS facility

GPS for vehicle

The ratio of vehicles with GPS functionality to total vehicle

[8]

5 Result and Discussion The average trip length for app-based bike taxi service in Guwahati is found to be 4.8 km. It is found that just 20% of trips are longer than 6 km and 50% are shorter than 4 km. It also observed that most of the trip makers in less than 5000 thousand income group are in 20–25 age group who use service primarily for college or any other work-related trips. The survey shows that the online fare system, door-to-door service, and online booking are major attractions for choosing app-based bike taxi services. The performance by both benchmarking and NRA is explained below.

5.1 Performance of Bike Taxi by Benchmarking Technique Each indicator considerably influences the app-based bike taxi service’s overall LOS. The calculated Level of Service (CLOS) for chosen indicators section is discussed below:

Quality Assessment of App-Based Bike Taxi Services by Benchmarking …

741

Extent of Supply: Lack of bike taxis will lead to users switching, while an uncontrolled supply of bike taxis may cause traffic issues and safety concerns. The presence of bike taxis/1000 population obtained 10.44, hence based on Table 2, CLOS is determined to be 4. Service Coverage: It measures how easily a service can be offered at different locations, more service coverage will result in better mode choice. The total length of the corridors on which app-based bike taxis ply in the city is 171.6 km. The area of the urban limits of the city is 216 km2 . The ratio of both gives the service coverage of app-based bike taxi service as 0.79. The respective CLOS is 2 based on Table 2. Comfort: The CLOS measure for comfort was taken to be the geometric mean of the relative values for the four constituent elements of comfort. The relative values of the driver behaviour, simplicity to use service, luggage carrying capacity, and fare collection system are 0.668, 0.656, 0.528, and 0.674, respectively. The geometric mean of the four attributes of comfort is 0.6284, which shows the overall acceptance is 62.84%. Thus, based on the LOS criteria of comfort in Table 2, CLOS is found to be 2. Average waiting time: The degree of waiting time is directly concerned with the reliability of the bike taxi service. The average waiting time calculated from a survey at different locations was found to be 6.16 min. Therefore, the CLOS for the average waiting time of app-based bike taxis is found to be 3 based on the LOS criteria. LOS criteria for the above indicators are given in Table 2. Travel speed Traffic congestion is reflected by travel speed. The average speed of app-based bike taxi service along major routes in the city is obtained to be 30.40 kmph, which indicates CLOS of app-based bike taxis for the city is 1. Affordability According to the trip maker survey, the average trip maker affordability across all income levels is 13.12%, and the CLOS is 2.

Table 2 LOS criteria for extent of supply, comfort, service coverage, and average waiting time LOS criteria Extent of supply Comfort (geometric Service coverage Average waiting time mean of relative (minutes) values) 1

0.85

≥1

8

Investigating the influence of income on commute distance > Examining the influence of socio-economic characteristics on commute distance > Identifying the influence of family size on commute distance > Scrutinizing the impact residential characteristics have on commute distance.

4 Methodology and Data Sources For addressing these objectives we follow a purely quantitative approach. We use the third level of the 76th round of NSSO on “Drinking Water, Sanitation and Housing conditions.” [The National Sample Survey Organization (NSSO) is a part of the Ministry of Statistics and Programme Implementation (Mospi), Government of India. NSSO has conducted nationwide sample surveys on various socio-economic aspects since 1950. These surveys are conducted in the form of rounds extending normally over a period of one year. The sample is selected through a multi-level random sampling and is considered representative of India’s population.

Limits to Commute: The Case of Indian Women

767

Gender is one of the key socio-demographic variables that can influence travel behaviour, but it is often the least understood. As the focus of the present analysis is to understand the determinants of commuting in women, hence, the households where the women are not participating in commuting are removed from the data. This leaves us with 40745 observations or 38.17% of the original sample. We begin our analysis by descriptive statistics. This provides a preliminary overview of general behaviour trends on mean travel distance. Subsequent to this analysis, we compare and contrast the commute pattern amongst males and females. This analysis is carried out at two levels namely at the aggregated level and at the disaggregated level, i.e., rural and urban. In addition, we compare the commute distance characteristics of urban and rural females. As the variable is not jointly distributed, we use the cosine similarity for this comparison. The data set is collated at the household level, i.e., individual characteristics like age, level of education etc., are not collected. Therefore, for the analysis, we use household level socioeconomic characteristics. This is in addition to the nature of the residential area and the tenure of the dwelling (Refer Table 1) for characteristics of the variables).

For setting-up the regression equation, we examine the cross-tabulated results between the socio-economic variables and commute trip distance. This analysis will help in understanding whether the socio-economic variables have an impact on commute distance. Subsequent to this analysis we setup the regression equation. Though the data is ordinal in nature, the categories are exclusive, and hence, we use the stratified regression. Our regression equation takes the form yi = di + bi xi + bi wi + bi z i

(1)

where yi is the commute distance category, d i is the dummy which represents the area type namely urban or rural, x i represents all the socio-economic variables including the characteristics of the dwelling and bi is the estimator. Table 1 Description of variables Variable

Nature

Level

Dependent

Categorical

8

Predictor

Rural/Urban

Dummy

2

Predictor

Religion

Categorical (with Buddhism as the calibrating level)

8

Social Group

Categorical (With Other Backward class as the 4 calibrating level)

Tenurial status

Categorical (employee quarter as the base level)

7

Area of residence

Categorical (with denoted slum as the base level)

4

Family size

Integer (with minimum value of 1)

Usual monthly consumer Integer (proxy for income) expenditure

768

N. Gosavi and N. S. G. Dittakavi

4.1 Is There a Stark Difference Between Male and Female Commute Patterns? As a primer to our analysis, we begin by examining the commute of men and women across the different distance buckets at an aggregate and disaggregate level, i.e., rural and urban (Refer Table 1). Additionally, we use descriptive statistics for understanding how far men and women commute. A cursory examination of the commute pattern reveals that more than half the commute trips are for a distance of less than five kilometres and that at an aggregate level ‘1–5 kms’ is the modal and median class for both men and women. At the disaggregated level we observe that for men this characteristic is preserved. Whereas for rural women, the modal class is ‘1–5 km’, while the median class is ‘